{ "startedUtc": "2026-10-06T13:06:19Z", "command": "run-all", "commandLine": "~/gvb-work/lanes/v7-final/src/GenericVectorBuilder.Bench/bin/Release/net10.0/GenericVectorBuilder.Bench.dll run-all --pipeline eshoponweb --queries golden --seed 701 --out ~/ForClaude/GenericVectorBuilder/bench-results", "machine": { "host": "bench-host", "cpu": "Intel(R) Xeon(R) CPU E5-1620 v3 @ 3.50GHz", "logicalCpus": 8, "ramGiB": 62.7, "gpu": "NVIDIA GeForce RTX 3060, 12288 MiB", "os": "Ubuntu 24.04.5 LTS, kernel 6.8.0-142-generic", "dotNet": ".NET 10.0.12", "loadAverage": "0.81 0.48 0.28" }, "pipeline": "eshoponweb", "collection": "gvbbench_eshoponweb", "rows": 524, "dimension": 1024, "source": "Read READ-ONLY from GenericVectorBuilder.dbo.gvb_eshoponweb in ChunkId order, vectors as native SqlVector\u003Cfloat\u003E, 0.2 s.", "queries": "golden: 20 labelled questions from ~/ForClaude/evalkit/questions_golden.json, embedded with qwen3-emb-0.6b (cached in ~/gvb-data/bench-cache/golden-eshoponweb-68df42ca69efa079.json, no embedding calls)", "queryCount": 20, "top": 10, "concurrency": [ 1, 8 ], "secondsPerLevel": 20, "truthSeconds": 0.0239995, "truthNdcg": 0.5181699774768911, "runSeed": 701, "targetOrder": [ "clickhouse", "vespa", "oracle", "elasticsearch", "sql", "pgvector", "sqlitevec", "qdrant", "opensearch", "qdrant-hnsw", "redis", "milvus", "mariadb", "weaviate", "mongodb", "chroma", "typesense", "sql-diskann", "duckdb" ], "notes": [ "Order: targets ran one at a time in a random order from seed 701 (runSeed; --seed 701 repeats it, targetOrder lists it). Inside each target the timed passes also ran in a random order from the same seed and the target\u0027s name (passOrder): default@1 is one searcher for 20 s (every search\u0027s latency gives p50/p95/p99, the completed searches give QPS@1, its first answer to each query gives recall and nDCG), default@N is N searchers for 20 s (QPS@N), exact is the engine\u0027s exact mode, one searcher for 60 s cycling the queries.", "Preparation, untimed, before any timed pass: a rehearsal of every pass type at its own concurrency for 30 s each, through the same code the passes use.", "Warm-up and settle check, untimed, right before every timed pass: the pass\u0027s own search at the pass\u0027s own number of searchers for at least 15 s and at least 20 searches (at most 120 s), read in windows of at least 2 s and 100 searches; then a 3 s trial of the same pass. The trial\u0027s figure (p50 with one searcher, QPS with several) must lie within 10% of the warm-up\u0027s settled figure (the median of its last 3 windows, which must agree within 5%); if not, the warm-up is extended once (at least 30 s, until its windows agree, at most 120 s), where an extension\u0027s windows agree when its level test passes: the older and the newer half of its latest windows, 5 to 10 windows a half, each half read as the pass reads it (the p50 of all its searches with one searcher, its searches per second with several), agree within 5%, judged on the windows that stopped it (an extension whose cap runs out with fewer than 10 windows is judged by its last 3 instead); then a second trial is taken, which must lie inside the range of the newer half\u0027s windows or within 10% of the newer half\u0027s figure, and the warm-up runs again before the pass. After the pass its own figure is held against the settled figure, within 10%. In the level test, the second trial and the hold, two one-searcher p50s no more than 0.1 ms apart agree whatever their percentage (below what this box resolves at one searcher). Each target\u0027s notes give every check, and a pass that still disagrees, or whose own figure did not hold, flags its target as unsettled. With machine control on, the check for a quiet box is made when the warm-up is announced, before it starts (a wait between the warm-up and the timed pass let the engine go cold), so by the time the clock opens that check is as old as the warm-up and the trial (about 18 s, more after an extension); each pass\u0027s conditions record that lead (quietCheckLeadSeconds). Failed untimed searches are counted per target (warmupErrors) and are not in the timed error counts.", "Load rows/s counts only time inside each target\u0027s upsert calls: one writer, batches of 1000, rows already in memory, the collection dropped and created fresh first. Engines that build or finish their index after the writes do it in a separate timed index step (the load\u0027s index seconds), and the run waits for it before searching.", "Index proof: each engine\u0027s own report of its index (indexState) is read after the load and again after the last pass. A target whose index was not ready after the load was still measured and carries a WARNING; an engine that reports nothing counts as not ready.", "Search settings: each searched target records the settings its own index description states (searchSettings: the build parameters and the effort per query, such as m, ef_construction and ef_search), read after the load; consolidating never averages runs whose settings differ.", "Durability: each target\u0027s crash-safety setting as configured here (durability). Engines that do not force writes to disk on every commit load faster for that reason.", "Latency is client-side wall time around each search (network and driver included), every search of the default@1 window, one searcher, the queries cycled; a window with fewer than 200 searches, a p50 above 1.25 x its mean, or a mean above its p99 is flagged.", "QPS: N searchers back to back for 20 s per level; completed searches divided by the window\u0027s elapsed time.", "Recall@10: share of the exact top 10 (brute force in memory) that the engine returned. A hit whose exact similarity ties the 10th best (within 1e-5) also counts, because duplicate rows embed to identical vectors.", "Routes: a container engine (the benchmark\u0027s own SQL Server and Qdrant containers included) is reached at its container\u0027s own address on its Docker network, never through the published localhost port (docker-proxy); the native comparison targets (sql-native, qdrant-native) are native services reached directly over loopback. Each target\u0027s addresses and the connections the client held open after its passes are in its notes and in conditions.connections.", "RAM of compose engines, the benchmark\u0027s own SQL Server and Qdrant containers (sql, sql-diskann, qdrant, qdrant-hnsw) included, is docker stats of the engine\u0027s containers; for the native comparison targets (sql-native, qdrant-native) it is the whole native process, including every other database or collection it serves. Disk is the table\u0027s reserved pages (SQL), the collection folder (Qdrant), or for other engines what the load added to the engine\u0027s data folder (engines that keep data in memory until a snapshot show almost nothing).", "Engines: run-all starts an engine that is down and stops it afterwards only if it was not running when the run began; an engine that was already running is left running.", "WARNING: the engine did not report a ready index after the load for: elasticsearch. They were measured anyway; their numbers may come from a scan or a half-built index (see each target\u0027s index state).", "Machine control on: governor performance on every CPU during the run (before: schedutil; at the end: performance; after putting it back: schedutil). CPU clock pinned for the run: turbo off (intel_pstate/no_turbo 0 -\u003E 1), so every CPU\u0027s clock is held at its ceiling of 3500 MHz whatever the engine runs; the uncore (L3 and memory) clock is held at 3000 MHz (min 3000 MHz, max 3000 MHz (MSR 0x620 = 0x1e1e); was min 1200 MHz, max 3000 MHz (MSR 0x620 = 0xc1e)) (before: turbo on (intel_pstate/no_turbo 0), ceiling 3600 MHz on CPUs 0-7, floor 1200 MHz on CPUs 0-7; while pinned: turbo off (intel_pstate/no_turbo 1), ceiling 3500 MHz on CPUs 0-7, floor 1200 MHz on CPUs 0-7; at the end: turbo off (intel_pstate/no_turbo 1), ceiling 3500 MHz on CPUs 0-7, floor 1200 MHz on CPUs 0-7; after putting it back: turbo on (intel_pstate/no_turbo 0), ceiling 3600 MHz on CPUs 0-7, floor 1200 MHz on CPUs 0-7). Uncore limit at the end: min 3000 MHz, max 3000 MHz (MSR 0x620 = 0x1e1e); after putting it back: min 1200 MHz, max 3000 MHz (MSR 0x620 = 0xc1e). Each target\u0027s clock note gives every pass\u0027s average MHz on the engine CPUs and on the client CPUs; a pass whose average on either is more than 1% off the pinned 3500 MHz is flagged. Figures from runs made with turbo on are not comparable with these in absolute terms. engine CPUs 2-3,6-7 (cores 2,6 and 3,7), client CPUs 0-1,4-5 (cores 0,4 and 1,5); the client process was pinned; each engine was pinned to the engine CPUs for its turn and put back after (conditions.engines); an engine run-all started (and stopped) was asked to be created on them, and its notes say whether the host did so or it was moved there after the start. Busy box: right before each pass\u0027s warm-up the run waits, up to 10 min, while processes outside the benchmark (everything but this client and the engine under test\u0027s cgroups) use more than 0.3 CPUs on average over the last 60 s or the last 5 s, neither window reaching back past the start of the target or of its engine (so the engine\u0027s own start-up is not outside work); if the box does not clear the pass runs anyway, flagged \u0027busy box\u0027, as is a pass whose own outside load is above the limit. Outside load counts busy = user \u002B nice \u002B system \u002B irq \u002B softirq \u002B steal (guest time is already in user); CONFIG_IRQ_TIME_ACCOUNTING is not set, so task and cgroup run time include the interrupt and softirq time that hit them and it is added; CONFIG_PARAVIRT_TIME_ACCOUNTING is not set, so steal is added (/boot/config-6.8.0-142-generic) (conditions.cpuAccounting); each pass also records this client\u0027s own CPU time per search (conditions.passes[].clientCpuMsPerSearch). CPU clocks were sampled every 250 ms; each pass\u0027s min/median/max per CPU is in conditions.passes. CPU idle states, recorded and left as found: driver intel_idle, governor menu, intel_idle max_cstate 9; POLL on, C1 on, C1E on, C3 OFF on CPUs 0-7 (default disabled), C6 on (conditions.cpuIdle). How each target was reached: conditions.connections. Client build Release, .NET 10.0.12.", "sqlitevec is embedded: it ran inside the client process on the client CPUs 0-1,4-5, sharing them with the client.", "duckdb is embedded: it ran inside the client process on the client CPUs 0-1,4-5, sharing them with the client.", "WARNING: busy box during mongodb exact: processes outside the benchmark used 0.31 CPUs on average during the pass, above the limit of 0.3.", "WARNING: clock off its pinned value during oracle default@8: the engine CPUs averaged 3121 MHz, -10.8% against the pinned 3500 MHz; the client CPUs averaged 3141 MHz, -10.3% against the pinned 3500 MHz (limit 1%); this pass is not comparable with passes at the pinned clock." ], "targets": [ { "name": "clickhouse", "engine": "ClickHouse 26.3.39.7 (MergeTree \u002B vector_similarity index; image-default system logging, each log table with a TTL of 1 hour)", "index": "vector_similarity HNSW cosineDistance, quantization bf16, M=16 ef_construction=128, hnsw_candidate_list_size_for_search=256, rescoring off; exact mode = full scan with skip indexes off", "hosting": "compose", "searchSettings": { "M": "16", "ef_construction": "128", "hnsw_candidate_list_size_for_search": "256" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.1639364, "rowsPerSecond": 3196.3615158073494, "countMatchSeconds": 0.0041315, "indexSeconds": 6.1884057, "indexNote": "1 data part(s), index files on 1 of them, 524 of 524 live rows in indexed parts; 0 merge(s) running, 0 unfinished mutation(s); plan of the default search uses the Skip index vec_idx on 1 of 1 parts; wait took 6.2 s" }, "search": { "countInTarget": 524, "latencySamples": 3832, "p50Ms": 4.97305, "p95Ms": 7.4002799999999995, "p99Ms": 8.143293, "qps": { "1": 191.56794302041496, "8": 374.7920878371932 }, "recall": 0.9949999999999999, "ndcg": 0.5276466892383727, "errors": 0, "exactQueries": 8686, "exactP50Ms": 6.67405, "exactP95Ms": 9.146675, "exactRecall": 1, "clientCpuMsPerSearch": { "1": 0.873, "8": 1.0567 } }, "ram": { "bytes": 937426944, "text": "894 MiB (docker stats)" }, "disk": { "bytes": 4515349223, "text": "4.21 GiB added by this load (4.21 GiB (whole engine data folder), 99.17 KiB before)" }, "passOrder": [ "default@1", "default@8", "exact" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "1 data part(s), index files on 1 of them, 524 of 524 live rows in indexed parts; 0 merge(s) running, 0 unfinished mutation(s); plan of the default search uses the Skip index vec_idx on 1 of 1 parts" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "1 data part(s), index files on 1 of them, 524 of 524 live rows in indexed parts; 0 merge(s) running, 0 unfinished mutation(s); plan of the default search uses the Skip index vec_idx on 1 of 1 parts" } }, "durability": "Acknowledged inserts are not fsynced. The sink creates plain MergeTree tables, so the MergeTree settings are the defaults: fsync_after_insert 0 and fsync_part_directory 0 (system.merge_tree_settings), and async_insert 1 with wait_for_async_insert 1 is the server default for the gvb login (system.settings), so an insert is acknowledged once its part is written, not once it is on disk. Measured 2026-10-04 with strace over 30 acknowledged single-row inserts (30 Ok rows in system.asynchronous_insert_log): zero fsync, fdatasync or sync_file_range calls, while the control, a table created with fsync_after_insert=1, made 61 fdatasync calls for 5 inserts. A host power loss or kernel crash can lose acknowledged rows that are still in the OS page cache; a ClickHouse process crash alone should not (inferred, not tested).", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started clickhouse.compose.yaml with every container created on CPUs 2-3,6-7: gvb-clickhouse cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [clickhouse-config/gvb-system-logs.xml 02cc53f0f171; clickhouse.compose.yaml 560e0ae487c6]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (1 data part(s), index files on 1 of them, 524 of 524 live rows in indexed parts; 0 merge(s) running, 0 unfinished mutation(s); plan of the default search uses the Skip index vec_idx on 1 of 1 parts). Durability: Acknowledged inserts are not fsynced. The sink creates plain MergeTree tables, so the MergeTree settings are the defaults: fsync_after_insert 0 and fsync_part_directory 0 (system.merge_tree_settings), and async_insert 1 with wait_for_async_insert 1 is the server default for the gvb login (system.settings), so an insert is acknowledged once its part is written, not once it is on disk. Measured 2026-10-04 with strace over 30 acknowledged single-row inserts (30 Ok rows in system.asynchronous_insert_log): zero fsync, fdatasync or sync_file_range calls, while the control, a table created with fsync_after_insert=1, made 61 fdatasync calls for 5 inserts. A host power loss or kernel crash can lose acknowledged rows that are still in the OS page cache; a ClickHouse process crash alone should not (inferred, not tested)..", "Outside load when searching began: 0.25 CPUs on average over the last 6 s and 0.26 over the last few seconds (limit 0.3), a quiet box. The load average 0.69 0.46 0.27 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@1 5,919 searches in 30.0 s, default@8 13,071 searches in 30.0 s, exact 4,223 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@1: warm-up 15.0 s and 2,956 searches (0 failed) at 1 searcher; p50 of the last windows 4.924, 4.968, 4.906 ms (windows of at least 2 s and 100 searches); trial of 602 searches in 3.0 s at 1 searcher: p50 4.901 ms against the settled p50 4.924 ms, 0% apart (limit 10%); the timed pass p50 4.973 ms, 1% from the settled p50 4.924 ms (limit 10% or 0.1 ms; 0.050 ms); default@8: warm-up 15.0 s and 6,536 searches (0 failed) at 8 searchers; QPS of the last windows 470, 464, 417 (windows of at least 2 s and 100 searches) (its last 3 windows differed by more than 5%); trial of 1,334 searches in 3.0 s at 8 searchers: 443 QPS against the settled 464 QPS, 5% apart (limit 10%), so the warm-up was EXTENDED once (32.1 s and 13,297 searches (0 failed) at 8 searchers; QPS of its older 7 windows 418 (windows 459, 327, 418, 466, 460, 435, 364) and of its newer 7 401 (windows 431, 450, 458, 391, 294, 364, 417), 4.4% apart (limit 5%) (windows of at least 2 s and 100 searches)); second trial of 1,153 searches in 3.0 s at 8 searchers: 383 QPS against the settled 401 QPS, 5% apart (limit 10%), inside its newer windows\u0027 range 294 to 458 QPS; settled after the extension; the timed pass 375 QPS, 7% from the settled 401 QPS (limit 10%); exact: warm-up 15.0 s and 1,921 searches (0 failed) at 1 searcher; p50 of the last windows 7.100, 6.871, 6.810 ms (windows of at least 2 s and 100 searches); trial of 389 searches in 3.0 s at 1 searcher: p50 6.936 ms against the settled p50 6.871 ms, 1% apart (limit 10%); the timed pass p50 6.674 ms, 3% from the settled p50 6.871 ms (limit 10% or 0.1 ms; 0.197 ms).", "Pass default@1 after rehearsal: 2026-10-06T13:08:22.784Z to 2026-10-06T13:08:42.787Z (20.0 s), 3,832 searches, 0 failed, p50 4.973 ms, mean 5.217 ms, p99 8.143 ms, 191.6 QPS (1000/QPS 5.220 ms).", "Pass default@8 after default@1: 2026-10-06T13:09:53.945Z to 2026-10-06T13:10:13.956Z (20.0 s), 7,500 searches, 0 failed, p50 20.844 ms, mean 21.335 ms, p99 39.445 ms, 374.8 QPS (1000/QPS 2.668 ms).", "Pass exact after default@8: 2026-10-06T13:10:31.966Z to 2026-10-06T13:11:31.968Z (60.0 s), 8,686 searches, 0 failed, p50 6.674 ms, mean 6.904 ms, p99 10.275 ms, 144.8 QPS (1000/QPS 6.908 ms).", "Passes in the order run: default@1, default@8, exact; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 57,827 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (1 data part(s), index files on 1 of them, 524 of 524 live rows in indexed parts; 0 merge(s) running, 0 unfinished mutation(s); plan of the default search uses the Skip index vec_idx on 1 of 1 parts).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-clickhouse (32cf2c26b272) on network gvb-clickhouse_default at container-ip:8123, not through docker-proxy localhost:8123. Open after the passes: localhost:1433 x2 (not this target), container-ip:8123 x1 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-clickhouse: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-clickhouse: 680 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@1 3492/3492 MHz, average 3492.0/3492.4, performance, outside load 0.24 before the warm-up, 0.25 during, client CPU 0.873 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.3, performance, outside load 0.25 before the warm-up after waiting 3 s, 0.25 during, client CPU 1.057 ms per search; exact 3492/3492 MHz, average 3491.9/3492.4, performance, outside load 0.25 before the warm-up, 0.25 during, client CPU 0.901 ms per search." ] }, { "name": "vespa", "engine": "Vespa 8.754.14 (tensor attribute \u002B HNSW)", "index": "HNSW float32 tensor, prenormalized-angular (cosine), max-links-per-node=16, neighbors-to-explore-at-insert=128; search targetHits=top, ef=100 via exploreAdditionalHits; exact mode = approximate:false; vectors held in memory", "hosting": "compose", "searchSettings": { "ef": "100", "max-links-per-node": "16", "neighbors-to-explore-at-insert": "128", "targetHits": "top" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 1.895675, "rowsPerSecond": 276.4186899125641, "countMatchSeconds": 0.0090197, "indexSeconds": 0.0519628, "indexNote": "nothing to build, Vespa inserts into the HNSW graph while writing; ready after 0.1 s: HNSW graph returns 524 of 524 vectors: totalCount 524, coverage full True, nearestNeighbor approximate:true with targetHits above the count: has_index True, algorithm \u0022index top k\u0022, top_k_hits 524; the graph is updated while each write is applied, so there is no later build step" }, "search": { "countInTarget": 524, "latencySamples": 9831, "p50Ms": 1.9701, "p95Ms": 2.4221, "p99Ms": 2.739120000000002, "qps": { "1": 491.5462200095681, "8": 1526.1773279584363 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 29472, "exactP50Ms": 1.9631, "exactP95Ms": 2.4515599999999975, "exactRecall": 1, "clientCpuMsPerSearch": { "1": 0.9579, "8": 0.8801 } }, "ram": { "bytes": 3151432253, "text": "2.93 GiB (docker stats)" }, "disk": { "bytes": 31064503, "text": "29.63 MiB added by this load (2.15 GiB (whole engine data folder), 2.12 GiB before)" }, "passOrder": [ "exact", "default@1", "default@8" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "HNSW graph returns 524 of 524 vectors: totalCount 524, coverage full True, nearestNeighbor approximate:true with targetHits above the count: has_index True, algorithm \u0022index top k\u0022, top_k_hits 524; the graph is updated while each write is applied, so there is no later build step" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "HNSW graph returns 524 of 524 vectors: totalCount 524, coverage full True, nearestNeighbor approximate:true with targetHits above the count: has_index True, algorithm \u0022index top k\u0022, top_k_hits 524; the graph is updated while each write is applied, so there is no later build step" } }, "durability": "Vespa\u0027s transaction log server fsyncs after each commit: searchlib.translogserver usefsync=true, and an operation is searchable when it is acknowledged: proton documentdb visibilitydelay=0. Neither is overridden by this sink or by the services.xml it deploys; VespaReadinessTests reads both from the live config server. After a crash the node replays the transaction log over its last flushed data and rebuilds the in-memory HNSW graph. By those settings an acknowledged write survives a crash; this is read from the settings, not shown by pulling power. One node and min-redundancy 1, so a lost disk loses the data.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started vespa.compose.yaml with every container created on CPUs 2-3,6-7: gvb-vespa cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [vespa.compose.yaml 3cfb702cc29f]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (HNSW graph returns 524 of 524 vectors: totalCount 524, coverage full True, nearestNeighbor approximate:true with targetHits above the count: has_index True, algorithm \u0022index top k\u0022, top_k_hits 524; the graph is updated while each write is applied, so there is no later build step). Durability: Vespa\u0027s transaction log server fsyncs after each commit: searchlib.translogserver usefsync=true, and an operation is searchable when it is acknowledged: proton documentdb visibilitydelay=0. Neither is overridden by this sink or by the services.xml it deploys; VespaReadinessTests reads both from the live config server. After a crash the node replays the transaction log over its last flushed data and rebuilds the in-memory HNSW graph. By those settings an acknowledged write survives a crash; this is read from the settings, not shown by pulling power. One node and min-redundancy 1, so a lost disk loses the data..", "Outside load when searching began: 0.23 CPUs on average over the last 40 s and 0.27 over the last few seconds (limit 0.3), a quiet box. The load average 6.39 4.62 2.28 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: exact 8,265 searches in 30.0 s, default@1 12,104 searches in 30.0 s, default@8 42,410 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. exact: warm-up 15.0 s and 6,825 searches (0 failed) at 1 searcher; p50 of the last windows 2.022, 2.025, 1.998 ms (windows of at least 2 s and 100 searches); trial of 1,455 searches in 3.0 s at 1 searcher: p50 1.970 ms against the settled p50 2.022 ms, 3% apart (limit 10%); the timed pass p50 1.963 ms, 3% from the settled p50 2.022 ms (limit 10% or 0.1 ms; 0.059 ms); default@1: warm-up 15.0 s and 7,383 searches (0 failed) at 1 searcher; p50 of the last windows 1.966, 1.986, 1.971 ms (windows of at least 2 s and 100 searches); trial of 1,475 searches in 3.0 s at 1 searcher: p50 1.971 ms against the settled p50 1.971 ms, 0% apart (limit 10%); the timed pass p50 1.970 ms, 0% from the settled p50 1.971 ms (limit 10% or 0.1 ms; 0.001 ms); default@8: warm-up 15.0 s and 22,144 searches (0 failed) at 8 searchers; QPS of the last windows 1,510, 1,483, 1,434 (windows of at least 2 s and 100 searches) (its last 3 windows differed by more than 5%); trial of 4,536 searches in 3.0 s at 8 searchers: 1,511 QPS against the settled 1,483 QPS, 2% apart (limit 10%), so the warm-up was EXTENDED once (30.0 s and 45,836 searches (0 failed) at 8 searchers; QPS of its older 6 windows 1,522 (windows 1,471, 1,537, 1,520, 1,542, 1,532, 1,532) and of its newer 6 1,529 (windows 1,540, 1,530, 1,528, 1,529, 1,507, 1,538), 0.4% apart (limit 5%) (windows of at least 2 s and 100 searches)); second trial of 4,502 searches in 3.0 s at 8 searchers: 1,499 QPS against the settled 1,529 QPS, 2% apart (limit 10%), outside its newer windows\u0027 range 1,507 to 1,540 QPS; settled after the extension; the timed pass 1,526 QPS, 0% from the settled 1,529 QPS (limit 10%).", "Pass exact after rehearsal: 2026-10-06T13:14:24.184Z to 2026-10-06T13:15:24.184Z (60.0 s), 29,472 searches, 0 failed, p50 1.963 ms, mean 2.033 ms, p99 2.839 ms, 491.2 QPS (1000/QPS 2.036 ms).", "Pass default@1 after exact: 2026-10-06T13:15:42.220Z to 2026-10-06T13:16:02.220Z (20.0 s), 9,831 searches, 0 failed, p50 1.970 ms, mean 2.031 ms, p99 2.739 ms, 491.5 QPS (1000/QPS 2.034 ms).", "Pass default@8 after default@1: 2026-10-06T13:17:08.249Z to 2026-10-06T13:17:28.253Z (20.0 s), 30,530 searches, 0 failed, p50 5.107 ms, mean 5.239 ms, p99 8.567 ms, 1526.2 QPS (1000/QPS 0.655 ms).", "Passes in the order run: exact, default@1, default@8; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 179,832 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (HNSW graph returns 524 of 524 vectors: totalCount 524, coverage full True, nearestNeighbor approximate:true with targetHits above the count: has_index True, algorithm \u0022index top k\u0022, top_k_hits 524; the graph is updated while each write is applied, so there is no later build step).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-vespa (d4d665486ac3) on network engines_default at container-ip:8080, not through docker-proxy localhost:8090; container gvb-vespa (d4d665486ac3) on network engines_default at container-ip:19071, not through docker-proxy localhost:19071. Open after the passes: container-ip:8080 x1 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-vespa: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-vespa: 215 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): exact 3492/3492 MHz, average 3492.0/3492.4, performance, outside load 0.22 before the warm-up, 0.19 during, client CPU 0.932 ms per search; default@1 3492/3492 MHz, average 3492.0/3492.3, performance, outside load 0.19 before the warm-up, 0.19 during, client CPU 0.958 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.1, performance, outside load 0.27 before the warm-up, 0.28 during, client CPU 0.880 ms per search." ] }, { "name": "oracle", "engine": "Oracle AI Database Free 23.26 (23ai line, VECTOR FLOAT32)", "index": "HNSW in-memory neighbor graph NEIGHBORS=16 EFCONSTRUCTION=128, EFSEARCH=100 per query, cosine; exact mode = FETCH EXACT FIRST (full scan); Oracle Free caps itself at 2 CPUs (cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit)", "hosting": "compose", "searchSettings": { "EFCONSTRUCTION": "128", "EFSEARCH": "100", "NEIGHBORS": "16" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.3030971, "rowsPerSecond": 1728.8189164462476, "countMatchSeconds": 0.0574364, "indexSeconds": 3.7880182, "indexNote": "HNSW graph holds 524 of 524 rows, change log waiting: 0 inserts and 0 deletes, USER_INDEXES status VALID, plan of the default search: VECTOR INDEX HNSW SCAN \u003E TABLE ACCESS BY INDEX ROWID, index used by 0 queries so far; Oracle Free caps itself at 2 CPUs (cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit); graph rebuilt in 2.5 s" }, "search": { "countInTarget": 524, "latencySamples": 21018, "p50Ms": 0.9269, "p95Ms": 1.1174149999999998, "p99Ms": 1.778614999999999, "qps": { "8": 2802.2354333862163, "1": 1050.894051939666 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 23504, "exactP50Ms": 2.4839, "exactP95Ms": 2.9561699999999997, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 0.6434, "1": 0.8802 } }, "ram": { "bytes": 2289217568, "text": "2.13 GiB (docker stats)" }, "disk": { "bytes": 0, "text": "0 B added by this load (6.9 GiB (whole engine data folder), 6.9 GiB before)" }, "passOrder": [ "default@8", "exact", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "HNSW graph holds 524 of 524 rows, change log waiting: 0 inserts and 0 deletes, USER_INDEXES status VALID, plan of the default search: VECTOR INDEX HNSW SCAN \u003E TABLE ACCESS BY INDEX ROWID, index used by 0 queries so far; Oracle Free caps itself at 2 CPUs (cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit)" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "HNSW graph holds 524 of 524 rows, change log waiting: 0 inserts and 0 deletes, USER_INDEXES status VALID, plan of the default search: VECTOR INDEX HNSW SCAN \u003E TABLE ACCESS BY INDEX ROWID, index used by 402297 queries so far; Oracle Free caps itself at 2 CPUs (cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit)" } }, "durability": "By these settings a committed row should survive a crash or power loss. The sink uses a plain COMMIT and commit_logging, commit_wait and commit_write are unset in the database, so every commit waits for its redo to be written. Measured 2026-10-04: 50 separate client commits raised V$SYSSTAT \u0027redo synch writes\u0027 by 55 (the 50 commits plus the CREATE and DROP of the probe table), and the log writer and the datafile writer hold their files open with O_DSYNC (open flags 02110002, filesystemio_options none). The database runs NOARCHIVELOG (V$DATABASE.LOG_MODE), so redo serves crash recovery only and there is no point-in-time restore. The HNSW graph lives in the 768 MB vector memory pool (oracle-init/01-vector-memory.sh) and is not the durable copy; the table is. Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked. CPU: Oracle Free caps itself at 2 CPUs (cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit).", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started oracle.compose.yaml with every container created on CPUs 2-3,6-7: gvb-oracle cpuset 2-3,6-7, gvb-oracle-seed cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [oracle-init/01-vector-memory.sh 6452e4e1953e; oracle.compose.yaml 2a7fd367012f]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (HNSW graph holds 524 of 524 rows, change log waiting: 0 inserts and 0 deletes, USER_INDEXES status VALID, plan of the default search: VECTOR INDEX HNSW SCAN \u003E TABLE ACCESS BY INDEX ROWID, index used by 0 queries so far; Oracle Free caps itself at 2 CPUs (cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit)). Durability: By these settings a committed row should survive a crash or power loss. The sink uses a plain COMMIT and commit_logging, commit_wait and commit_write are unset in the database, so every commit waits for its redo to be written. Measured 2026-10-04: 50 separate client commits raised V$SYSSTAT \u0027redo synch writes\u0027 by 55 (the 50 commits plus the CREATE and DROP of the probe table), and the log writer and the datafile writer hold their files open with O_DSYNC (open flags 02110002, filesystemio_options none). The database runs NOARCHIVELOG (V$DATABASE.LOG_MODE), so redo serves crash recovery only and there is no point-in-time restore. The HNSW graph lives in the 768 MB vector memory pool (oracle-init/01-vector-memory.sh) and is not the durable copy; the table is. Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked. CPU: Oracle Free caps itself at 2 CPUs (measured 2026-10-04: cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit; this sink reads the live value when the index state is read)..", "Outside load when searching began: 0.20 CPUs on average over the last 6 s and 0.19 over the last few seconds (limit 0.3), a quiet box. The load average 8.17 6.30 3.60 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 85,640 searches in 30.4 s, exact 11,735 searches in 30.0 s, default@1 31,474 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 44,230 searches (0 failed) at 8 searchers; QPS of the last windows 2,358, 3,428, 2,526 (windows of at least 2 s and 100 searches) (its last 3 windows differed by more than 5%); trial of 8,682 searches in 3.0 s at 8 searchers: 2,893 QPS against the settled 2,526 QPS, 13% apart (limit 10%), so the warm-up was EXTENDED once (30.0 s and 84,503 searches (0 failed) at 8 searchers; QPS of its older 6 windows 2,890 (windows 3,334, 2,378, 3,320, 2,572, 2,837, 2,900) and of its newer 6 2,826 (windows 2,500, 3,336, 2,408, 3,287, 2,550, 2,878), 2.2% apart (limit 5%) (windows of at least 2 s and 100 searches)); second trial of 9,082 searches in 3.0 s at 8 searchers: 3,026 QPS against the settled 2,826 QPS, 7% apart (limit 10%), inside its newer windows\u0027 range 2,408 to 3,336 QPS; settled after the extension; the timed pass 2,802 QPS, 1% from the settled 2,826 QPS (limit 10%); exact: warm-up 15.0 s and 5,863 searches (0 failed) at 1 searcher; p50 of the last windows 2.509, 2.469, 2.491 ms (windows of at least 2 s and 100 searches); trial of 1,183 searches in 3.0 s at 1 searcher: p50 2.473 ms against the settled p50 2.491 ms, 1% apart (limit 10%); the timed pass p50 2.484 ms, 0% from the settled p50 2.491 ms (limit 10% or 0.1 ms; 0.007 ms); default@1: warm-up 15.0 s and 15,780 searches (0 failed) at 1 searcher; p50 of the last windows 0.916, 0.916, 0.927 ms (windows of at least 2 s and 100 searches); trial of 3,143 searches in 3.0 s at 1 searcher: p50 0.930 ms against the settled p50 0.916 ms, 1% apart (limit 10%); the timed pass p50 0.927 ms, 1% from the settled p50 0.916 ms (limit 10% or 0.1 ms; 0.011 ms).", "Pass default@8 after rehearsal: 2026-10-06T13:20:32.534Z to 2026-10-06T13:20:52.657Z (20.1 s), 56,389 searches, 0 failed, p50 1.637 ms, mean 2.852 ms, p99 3.554 ms, 2802.2 QPS (1000/QPS 0.357 ms).", "Pass exact after default@8: 2026-10-06T13:21:10.680Z to 2026-10-06T13:22:10.681Z (60.0 s), 23,504 searches, 0 failed, p50 2.484 ms, mean 2.550 ms, p99 3.863 ms, 391.7 QPS (1000/QPS 2.553 ms).", "Pass default@1 after exact: 2026-10-06T13:22:28.708Z to 2026-10-06T13:22:48.708Z (20.0 s), 21,018 searches, 0 failed, p50 0.927 ms, mean 0.950 ms, p99 1.779 ms, 1050.9 QPS (1000/QPS 0.952 ms).", "Passes in the order run: default@8, exact, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 343,671 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (HNSW graph holds 524 of 524 rows, change log waiting: 0 inserts and 0 deletes, USER_INDEXES status VALID, plan of the default search: VECTOR INDEX HNSW SCAN \u003E TABLE ACCESS BY INDEX ROWID, index used by 402297 queries so far; Oracle Free caps itself at 2 CPUs (cpu_count 2 in V$PARAMETER, edition FREE in V$INSTANCE, 8 host CPUs in V$OSSTAT NUM_CPUS; the 2 CPU thread limit is Oracle\u0027s documented Free edition limit)).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-oracle (0da3dc27bd31) on network gvb-oracle_default at container-ip:1521, not through docker-proxy localhost:1521. Open after the passes: container-ip:1521 x13 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-oracle: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-oracle: 92 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3121.3/3141.0, performance, outside load 0.26 before the warm-up, 0.22 during, client CPU 0.643 ms per search; exact 3492/3492 MHz, average 3492.2/3492.3, performance, outside load 0.26 before the warm-up, 0.23 during, client CPU 0.929 ms per search; default@1 3492/3492 MHz, average 3492.2/3491.9, performance, outside load 0.23 before the warm-up, 0.23 during, client CPU 0.880 ms per search.", "WARNING: clock off its pinned value during oracle default@8: the engine CPUs averaged 3121 MHz, -10.8% against the pinned 3500 MHz; the client CPUs averaged 3141 MHz, -10.3% against the pinned 3500 MHz (limit 1%); this pass is not comparable with passes at the pinned clock." ] }, { "name": "elasticsearch", "engine": "Elasticsearch 9.5.3 (dense_vector)", "index": "HNSW float32, no quantization, m=16, ef_construction=128, cosine; search k=top, num_candidates=100; 1 shard, 0 replicas; force-merged to one segment after the load (at 1,024 dimensions a segment under 1,043 vectors gets no graph)", "hosting": "compose", "searchSettings": { "ef_construction": "128", "k": "top", "m": "16", "num_candidates": "100" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 1.1601699, "rowsPerSecond": 451.6579856105558, "countMatchSeconds": 0.1161943, "indexSeconds": 0.3904655, "indexNote": "FAILED after 0.4 s: NO HNSW GRAPH, searches scan all 524 vectors: 1 segment(s), 524 vectors, total_vex_size_bytes 0 (_stats dense_vector), index_options hnsw m=16 ef_construction=128; Lucene builds no graph for a segment this small (measured at 1,024 dimensions: 1,042 vectors none, 1,043 vectors a graph), so default search equals exact search" }, "search": { "countInTarget": 524, "latencySamples": 14790, "p50Ms": 1.3287, "p95Ms": 1.5088549999999998, "p99Ms": 1.775022, "qps": { "8": 2671.293756146442, "1": 739.4510557346209 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 46096, "exactP50Ms": 1.2776, "exactP95Ms": 1.479825, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 0.4968, "1": 0.4632 } }, "ram": { "bytes": 2878701830, "text": "2.68 GiB (docker stats)" }, "disk": { "bytes": 2502299, "text": "2.39 MiB added by this load (2.56 MiB (whole engine data folder), 172.81 KiB before)" }, "passOrder": [ "exact", "default@8", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": false, "indexedVectors": 0, "totalVectors": 524, "detail": "NO HNSW GRAPH, searches scan all 524 vectors: 1 segment(s), 524 vectors, total_vex_size_bytes 0 (_stats dense_vector), index_options hnsw m=16 ef_construction=128; Lucene builds no graph for a segment this small (measured at 1,024 dimensions: 1,042 vectors none, 1,043 vectors a graph), so default search equals exact search" }, "afterSearch": { "ready": false, "indexedVectors": 0, "totalVectors": 524, "detail": "NO HNSW GRAPH, searches scan all 524 vectors: 1 segment(s), 524 vectors, total_vex_size_bytes 0 (_stats dense_vector), index_options hnsw m=16 ef_construction=128; Lucene builds no graph for a segment this small (measured at 1,024 dimensions: 1,042 vectors none, 1,043 vectors a graph), so default search equals exact search" } }, "durability": "Every acknowledged bulk request is fsynced to the translog before the answer: index.translog.durability=request, the Elasticsearch default, which neither this sink nor elasticsearch.compose.yaml overrides (ElasticsearchReadinessTests reads it back from the live index). By that setting a process crash or power loss loses no acknowledged write; this is read from the setting, not shown by pulling power. One node and no replicas, so a lost disk loses the data.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started elasticsearch.compose.yaml with every container created on CPUs 2-3,6-7: gvb-elasticsearch cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [elasticsearch.compose.yaml 582b9a97b8de]. This run created the engine from these files.", "WARNING: the index step after the load FAILED (NO HNSW GRAPH, searches scan all 524 vectors: 1 segment(s), 524 vectors, total_vex_size_bytes 0 (_stats dense_vector), index_options hnsw m=16 ef_construction=128; Lucene builds no graph for a segment this small (measured at 1,024 dimensions: 1,042 vectors none, 1,043 vectors a graph), so default search equals exact search). Searched anyway; see the index state.", "WARNING: index not ready after the load: NOT ready, 0 of 524 indexed (NO HNSW GRAPH, searches scan all 524 vectors: 1 segment(s), 524 vectors, total_vex_size_bytes 0 (_stats dense_vector), index_options hnsw m=16 ef_construction=128; Lucene builds no graph for a segment this small (measured at 1,024 dimensions: 1,042 vectors none, 1,043 vectors a graph), so default search equals exact search). Measured anyway, so its search numbers may come from a scan or a half-built index. Durability: Every acknowledged bulk request is fsynced to the translog before the answer: index.translog.durability=request, the Elasticsearch default, which neither this sink nor elasticsearch.compose.yaml overrides (ElasticsearchReadinessTests reads it back from the live index). By that setting a process crash or power loss loses no acknowledged write; this is read from the setting, not shown by pulling power. One node and no replicas, so a lost disk loses the data..", "Outside load when searching began: 0.19 CPUs on average over the last 2 s and 0.19 over the last few seconds (limit 0.3), a quiet box. The load average 2.32 4.36 3.61 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: exact 14,440 searches in 30.0 s, default@8 63,152 searches in 30.0 s, default@1 22,164 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. exact: warm-up 15.0 s and 11,249 searches (0 failed) at 1 searcher; p50 of the last windows 1.281, 1.282, 1.277 ms (windows of at least 2 s and 100 searches); trial of 2,298 searches in 3.0 s at 1 searcher: p50 1.281 ms against the settled p50 1.281 ms, 0% apart (limit 10%); the timed pass p50 1.278 ms, 0% from the settled p50 1.281 ms (limit 10% or 0.1 ms; 0.003 ms); default@8: warm-up 15.0 s and 39,834 searches (0 failed) at 8 searchers; QPS of the last windows 2,676, 2,636, 2,696 (windows of at least 2 s and 100 searches); trial of 7,942 searches in 3.0 s at 8 searchers: 2,646 QPS against the settled 2,676 QPS, 1% apart (limit 10%); the timed pass 2,671 QPS, 0% from the settled 2,676 QPS (limit 10%); default@1: warm-up 15.0 s and 11,065 searches (0 failed) at 1 searcher; p50 of the last windows 1.326, 1.329, 1.330 ms (windows of at least 2 s and 100 searches); trial of 2,200 searches in 3.0 s at 1 searcher: p50 1.335 ms against the settled p50 1.329 ms, 0% apart (limit 10%); the timed pass p50 1.329 ms, 0% from the settled p50 1.329 ms (limit 10% or 0.1 ms; 0.001 ms).", "Pass exact after rehearsal: 2026-10-06T13:25:15.144Z to 2026-10-06T13:26:15.144Z (60.0 s), 46,096 searches, 0 failed, p50 1.278 ms, mean 1.300 ms, p99 1.715 ms, 768.3 QPS (1000/QPS 1.302 ms).", "Pass default@8 after exact: 2026-10-06T13:26:33.198Z to 2026-10-06T13:26:53.200Z (20.0 s), 53,431 searches, 0 failed, p50 2.865 ms, mean 2.992 ms, p99 5.422 ms, 2671.3 QPS (1000/QPS 0.374 ms).", "Pass default@1 after default@8: 2026-10-06T13:27:11.230Z to 2026-10-06T13:27:31.232Z (20.0 s), 14,790 searches, 0 failed, p50 1.329 ms, mean 1.350 ms, p99 1.775 ms, 739.5 QPS (1000/QPS 1.352 ms).", "Passes in the order run: exact, default@8, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 174,344 sent, 0 failed (warmupErrors). Index after the searches: NOT ready, 0 of 524 indexed (NO HNSW GRAPH, searches scan all 524 vectors: 1 segment(s), 524 vectors, total_vex_size_bytes 0 (_stats dense_vector), index_options hnsw m=16 ef_construction=128; Lucene builds no graph for a segment this small (measured at 1,024 dimensions: 1,042 vectors none, 1,043 vectors a graph), so default search equals exact search).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-elasticsearch (f2716e40a93c) on network engines_default at container-ip:9200, not through docker-proxy localhost:9200. Open after the passes: container-ip:9200 x8 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-elasticsearch: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-elasticsearch: 75 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): exact 3492/3492 MHz, average 3491.9/3491.6, performance, outside load 0.22 before the warm-up, 0.22 during, client CPU 0.461 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.1, performance, outside load 0.22 before the warm-up, 0.24 during, client CPU 0.497 ms per search; default@1 3492/3492 MHz, average 3492.0/3492.3, performance, outside load 0.24 before the warm-up, 0.21 during, client CPU 0.463 ms per search." ] }, { "name": "sql", "engine": "Microsoft SQL Server 2025 (RTM-CU9) (KB5122048) - 17.0.5005.3 (X64), Enterprise Developer Edition (64-bit), in container gvb-mssql (image mcr.microsoft.com/mssql/server:2025-CU9-ubuntu-24.04@sha256:2b5b581621126574f3d1f75e78d3eebe8d05aedb59ad0cfdf9aa42cb0634d726; cpuset 2-3,6-7 from its creation; SQL Server counts 4 CPU(s) and runs 4 visible scheduler(s), affinity AUTO)", "index": "exact VECTOR_DISTANCE cosine, no vector index (full scan)", "hosting": "compose", "searchSettings": { "description": "exact VECTOR_DISTANCE cosine, no vector index (full scan)" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.8934161, "rowsPerSecond": 586.5128242036382, "countMatchSeconds": 0.0045552 }, "search": { "countInTarget": 524, "latencySamples": 4820, "p50Ms": 4.06025, "p95Ms": 4.58204, "p99Ms": 6.061826000000006, "qps": { "8": 611.6123746788679, "1": 240.98674331925002 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 0, "clientCpuMsPerSearch": { "8": 1.1838, "1": 1.2193 } }, "ram": { "bytes": 555640422, "text": "529.9 MiB (docker stats)" }, "disk": { "bytes": 5398528, "text": "5.15 MiB (table and its indexes, reserved pages)" }, "passOrder": [ "default@8", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 0, "totalVectors": 524, "detail": "no index used, exact scan by design: dbo.gvb_gvbbench_eshoponweb in GvbBench holds 524 rows; sys.vector_indexes lists no index on the table; indexes: PK_gvb_gvbbench_eshoponweb CLUSTERED, IX_gvb_gvbbench_eshoponweb_DocKey NONCLUSTERED; a real search under SET STATISTICS XML ON ran: Clustered Index Scan of GvbBench.gvb_gvbbench_eshoponweb through PK_gvb_gvbbench_eshoponweb (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits" }, "afterSearch": { "ready": true, "indexedVectors": 0, "totalVectors": 524, "detail": "no index used, exact scan by design: dbo.gvb_gvbbench_eshoponweb in GvbBench holds 524 rows; sys.vector_indexes lists no index on the table; indexes: PK_gvb_gvbbench_eshoponweb CLUSTERED, IX_gvb_gvbbench_eshoponweb_DocKey NONCLUSTERED; a real search under SET STATISTICS XML ON ran: Clustered Index Scan of GvbBench.gvb_gvbbench_eshoponweb through PK_gvb_gvbbench_eshoponweb (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits" } }, "durability": "A commit returns after its transaction-log records are written to disk (SQL Server write-ahead logging; delayed durability is DISABLED); a database created here copies the model database: recovery model FULL, page_verify CHECKSUM; no global trace flags are enabled; mssql.conf of container gvb-mssql (/var/opt/mssql/mssql.conf, on the host at ~/gvb-data/engines/mssql/mssql.conf) does not exist, so it sets: nothing; it has no [control] or [traceflag] entry, so SQL Server\u0027s own Linux defaults for flushing writes apply (Microsoft\u0027s Linux performance guide names trace flag 3982 as that default; read from the guide, not tested here). Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked. Container settings from its environment (names only): MSSQL_AGENT_ENABLED, MSSQL_MEMORY_LIMIT_MB, MSSQL_PID, MSSQL_RPC_PORT, MSSQL_SA_PASSWORD.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started mssql.compose.yaml with every container created on CPUs 2-3,6-7: gvb-mssql cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [mssql.compose.yaml b7359663cc15]. This run created the engine from these files.", "Index after the load: ready, 0 of 524 indexed (no index used, exact scan by design: dbo.gvb_gvbbench_eshoponweb in GvbBench holds 524 rows; sys.vector_indexes lists no index on the table; indexes: PK_gvb_gvbbench_eshoponweb CLUSTERED, IX_gvb_gvbbench_eshoponweb_DocKey NONCLUSTERED; a real search under SET STATISTICS XML ON ran: Clustered Index Scan of GvbBench.gvb_gvbbench_eshoponweb through PK_gvb_gvbbench_eshoponweb (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits). Durability: A commit returns after its transaction-log records are written to disk (SQL Server write-ahead logging; delayed durability is DISABLED); a database created here copies the model database: recovery model FULL, page_verify CHECKSUM; no global trace flags are enabled; mssql.conf of container gvb-mssql (/var/opt/mssql/mssql.conf, on the host at ~/gvb-data/engines/mssql/mssql.conf) does not exist, so it sets: nothing; it has no [control] or [traceflag] entry, so SQL Server\u0027s own Linux defaults for flushing writes apply (Microsoft\u0027s Linux performance guide names trace flag 3982 as that default; read from the guide, not tested here). Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked. Container settings from its environment (names only): MSSQL_AGENT_ENABLED, MSSQL_MEMORY_LIMIT_MB, MSSQL_PID, MSSQL_RPC_PORT, MSSQL_SA_PASSWORD..", "Outside load when searching began: 0.22 CPUs on average over the last 3 s and 0.22 over the last few seconds (limit 0.3), a quiet box. The load average 3.27 3.92 3.61 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 17,985 searches in 30.0 s, default@1 7,216 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 9,257 searches (0 failed) at 8 searchers; QPS of the last windows 625, 601, 622 (windows of at least 2 s and 100 searches); trial of 1,828 searches in 3.0 s at 8 searchers: 608 QPS against the settled 622 QPS, 2% apart (limit 10%); the timed pass 612 QPS, 2% from the settled 622 QPS (limit 10%); default@1: warm-up 15.0 s and 3,604 searches (0 failed) at 1 searcher; p50 of the last windows 4.078, 4.054, 4.051 ms (windows of at least 2 s and 100 searches); trial of 723 searches in 3.0 s at 1 searcher: p50 4.080 ms against the settled p50 4.054 ms, 1% apart (limit 10%); the timed pass p50 4.060 ms, 0% from the settled p50 4.054 ms (limit 10% or 0.1 ms; 0.006 ms).", "Pass default@8 after rehearsal: 2026-10-06T13:29:04.896Z to 2026-10-06T13:29:24.907Z (20.0 s), 12,239 searches, 0 failed, p50 12.650 ms, mean 13.074 ms, p99 19.890 ms, 611.6 QPS (1000/QPS 1.635 ms).", "Pass default@1 after default@8: 2026-10-06T13:29:42.916Z to 2026-10-06T13:30:02.917Z (20.0 s), 4,820 searches, 0 failed, p50 4.060 ms, mean 4.146 ms, p99 6.062 ms, 241.0 QPS (1000/QPS 4.150 ms).", "Passes in the order run: default@8, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 40,613 sent, 0 failed (warmupErrors). Index after the searches: ready, 0 of 524 indexed (no index used, exact scan by design: dbo.gvb_gvbbench_eshoponweb in GvbBench holds 524 rows; sys.vector_indexes lists no index on the table; indexes: PK_gvb_gvbbench_eshoponweb CLUSTERED, IX_gvb_gvbbench_eshoponweb_DocKey NONCLUSTERED; a real search under SET STATISTICS XML ON ran: Clustered Index Scan of GvbBench.gvb_gvbbench_eshoponweb through PK_gvb_gvbbench_eshoponweb (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-mssql (5e402e3af1a6) on network gvb-mssql_default at container-ip:1433, not through docker-proxy localhost:14330. Open after the passes: container-ip:1433 x9 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-mssql: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-mssql: 86 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3492.0/3492.0, performance, outside load 0.22 before the warm-up, 0.26 during, client CPU 1.184 ms per search; default@1 3492/3492 MHz, average 3492.4/3492.5, performance, outside load 0.23 before the warm-up, 0.21 during, client CPU 1.219 ms per search." ] }, { "name": "pgvector", "engine": "PostgreSQL 17 \u002B pgvector 0.8.7", "index": "HNSW vector_cosine_ops m=16 ef_construction=128, hnsw.ef_search=100 per query, float32 vector(n), cosine; exact mode = same query with index scans off (sequential scan)", "hosting": "compose", "searchSettings": { "ef_construction": "128", "hnsw.ef_search": "100", "m": "16" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.8654389, "rowsPerSecond": 605.4731304543856, "countMatchSeconds": 0.0118084, "indexSeconds": 0.0279119, "indexNote": "HNSW is maintained inside every insert, nothing to build; confirmed in 0.0 s: pg_indexes lists gvb_gvbbench_eshoponweb_hnsw (USING hnsw (embedding vector_cosine_ops) WITH (m=\u002716\u0027, ef_construction=\u0027128\u0027)), pg_index valid and ready = True, EXPLAIN of the default search uses Index Scan on it = True, that search returned 10 of 10 rows; PostgreSQL keeps no entry count for an HNSW index, so indexed vectors are not reported" }, "search": { "countInTarget": 524, "latencySamples": 22143, "p50Ms": 0.8846, "p95Ms": 1.0411899999999996, "p99Ms": 1.3103219999999982, "qps": { "8": 3933.8796240634597, "1": 1107.11890103007 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 26025, "exactP50Ms": 2.2571, "exactP95Ms": 2.56478, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 0.5376, "1": 0.6878 } }, "ram": { "bytes": 105381888, "text": "100.5 MiB (docker stats)" }, "disk": { "bytes": 7938048, "text": "7.57 MiB added by this load (374.57 MiB (whole engine data folder), 366.99 MiB before)" }, "passOrder": [ "exact", "default@8", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "totalVectors": 524, "detail": "pg_indexes lists gvb_gvbbench_eshoponweb_hnsw (USING hnsw (embedding vector_cosine_ops) WITH (m=\u002716\u0027, ef_construction=\u0027128\u0027)), pg_index valid and ready = True, EXPLAIN of the default search uses Index Scan on it = True, that search returned 10 of 10 rows; PostgreSQL keeps no entry count for an HNSW index, so indexed vectors are not reported" }, "afterSearch": { "ready": true, "totalVectors": 524, "detail": "pg_indexes lists gvb_gvbbench_eshoponweb_hnsw (USING hnsw (embedding vector_cosine_ops) WITH (m=\u002716\u0027, ef_construction=\u0027128\u0027)), pg_index valid and ready = True, EXPLAIN of the default search uses Index Scan on it = True, that search returned 10 of 10 rows; PostgreSQL keeps no entry count for an HNSW index, so indexed vectors are not reported" } }, "durability": "fsync on, synchronous_commit on, full_page_writes on, wal_sync_method fdatasync (PostgreSQL defaults; pgvector.compose.yaml sets only shared_buffers, maintenance_work_mem and max_wal_size): every commit is flushed to the write-ahead log before it returns and HNSW index changes are WAL-logged, so a crash loses no committed row (max_wal_size 4GB only spaces out checkpoints; measured with docker kill, which is SIGKILL, right after a 2,000-vector load and a restart: all 2,000 rows were there and the HNSW index was valid and used by the default search)", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started pgvector.compose.yaml with every container created on CPUs 2-3,6-7: gvb-pgvector cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [pgvector.compose.yaml 27549563c622]. This run created the engine from these files.", "Index after the load: ready, ? of 524 indexed (pg_indexes lists gvb_gvbbench_eshoponweb_hnsw (USING hnsw (embedding vector_cosine_ops) WITH (m=\u002716\u0027, ef_construction=\u0027128\u0027)), pg_index valid and ready = True, EXPLAIN of the default search uses Index Scan on it = True, that search returned 10 of 10 rows; PostgreSQL keeps no entry count for an HNSW index, so indexed vectors are not reported). Durability: fsync on, synchronous_commit on, full_page_writes on, wal_sync_method fdatasync (PostgreSQL defaults; pgvector.compose.yaml sets only shared_buffers, maintenance_work_mem and max_wal_size): every commit is flushed to the write-ahead log before it returns and HNSW index changes are WAL-logged, so a crash loses no committed row (max_wal_size 4GB only spaces out checkpoints; measured with docker kill, which is SIGKILL, right after a 2,000-vector load and a restart: all 2,000 rows were there and the HNSW index was valid and used by the default search).", "Outside load when searching began: 0.24 CPUs on average over the last 1 s and 0.24 over the last few seconds (limit 0.3), a quiet box. The load average 2.73 3.74 3.61 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: exact 13,387 searches in 30.0 s, default@8 122,772 searches in 30.0 s, default@1 33,961 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. exact: warm-up 15.0 s and 6,509 searches (0 failed) at 1 searcher; p50 of the last windows 2.226, 2.238, 2.264 ms (windows of at least 2 s and 100 searches); trial of 1,307 searches in 3.0 s at 1 searcher: p50 2.245 ms against the settled p50 2.238 ms, 0% apart (limit 10%); the timed pass p50 2.257 ms, 1% from the settled p50 2.238 ms (limit 10% or 0.1 ms; 0.019 ms); default@8: warm-up 15.0 s and 58,882 searches (0 failed) at 8 searchers; QPS of the last windows 3,900, 3,966, 3,916 (windows of at least 2 s and 100 searches); trial of 11,879 searches in 3.0 s at 8 searchers: 3,958 QPS against the settled 3,916 QPS, 1% apart (limit 10%); the timed pass 3,934 QPS, 0% from the settled 3,916 QPS (limit 10%); default@1: warm-up 15.0 s and 16,787 searches (0 failed) at 1 searcher; p50 of the last windows 0.868, 0.868, 0.878 ms (windows of at least 2 s and 100 searches); trial of 3,328 searches in 3.0 s at 1 searcher: p50 0.885 ms against the settled p50 0.868 ms, 2% apart (limit 10%); the timed pass p50 0.885 ms, 2% from the settled p50 0.868 ms (limit 10% or 0.1 ms; 0.016 ms).", "Pass exact after rehearsal: 2026-10-06T13:32:00.427Z to 2026-10-06T13:33:00.427Z (60.0 s), 26,025 searches, 0 failed, p50 2.257 ms, mean 2.303 ms, p99 3.157 ms, 433.7 QPS (1000/QPS 2.305 ms).", "Pass default@8 after exact: 2026-10-06T13:33:18.457Z to 2026-10-06T13:33:38.458Z (20.0 s), 78,682 searches, 0 failed, p50 1.985 ms, mean 2.032 ms, p99 3.545 ms, 3933.9 QPS (1000/QPS 0.254 ms).", "Pass default@1 after default@8: 2026-10-06T13:33:56.492Z to 2026-10-06T13:34:16.492Z (20.0 s), 22,143 searches, 0 failed, p50 0.885 ms, mean 0.901 ms, p99 1.310 ms, 1107.1 QPS (1000/QPS 0.903 ms).", "Passes in the order run: exact, default@8, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 268,812 sent, 0 failed (warmupErrors). Index after the searches: ready, ? of 524 indexed (pg_indexes lists gvb_gvbbench_eshoponweb_hnsw (USING hnsw (embedding vector_cosine_ops) WITH (m=\u002716\u0027, ef_construction=\u0027128\u0027)), pg_index valid and ready = True, EXPLAIN of the default search uses Index Scan on it = True, that search returned 10 of 10 rows; PostgreSQL keeps no entry count for an HNSW index, so indexed vectors are not reported).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-pgvector (6d6073de90e0) on network gvb-pgvector_default at container-ip:5432, not through docker-proxy localhost:5432. Open after the passes: container-ip:5432 x8 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-pgvector: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-pgvector: 6 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): exact 3492/3492 MHz, average 3492.4/3492.4, performance, outside load 0.22 before the warm-up, 0.20 during, client CPU 0.692 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.0, performance, outside load 0.20 before the warm-up, 0.27 during, client CPU 0.538 ms per search; default@1 3492/3492 MHz, average 3492.5/3491.9, performance, outside load 0.24 before the warm-up, 0.18 during, client CPU 0.688 ms per search." ] }, { "name": "sqlitevec", "engine": "SQLite 3.53.3 \u002B sqlite-vec 0.1.7-alpha.2.1 (vec0, embedded, in process)", "index": "vec0 brute-force scan, no ANN index (exact), float32, cosine distance, default chunk_size=1024; score = 1 - cosine distance; searches run concurrently, one WAL reader connection per searcher (opened as searchers arrive, at most 32), writes run one at a time and may overlap searches", "hosting": "embedded", "searchSettings": { "chunk_size": "1024" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.1285191, "rowsPerSecond": 4077.2149820532513, "countMatchSeconds": 0.0029446, "indexSeconds": 0.0041449, "indexNote": "nothing to build, every search scans all vectors; checked in 0.00 s: no index, exact scan by design: gvb_gvbbench_eshoponweb is a vec0 virtual table holding 524 vectors and every search compares all of them (EXPLAIN QUERY PLAN: SCAN gvb_gvbbench_eshoponweb VIRTUAL TABLE INDEX 0:3{___}___ | USE TEMP B-TREE FOR ORDER BY)" }, "search": { "countInTarget": 524, "latencySamples": 10198, "p50Ms": 1.9247, "p95Ms": 2.1821450000000002, "p99Ms": 2.342002000000002, "qps": { "1": 509.88742362669626, "8": 1141.9354931272935 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 30824, "exactP50Ms": 1.9056, "exactP95Ms": 2.16437, "exactRecall": 1, "clientCpuMsPerSearch": { "1": 2.0672, "8": 3.4925 } }, "ram": { "text": "in this process, not measured" }, "disk": { "bytes": 10196088, "text": "9.72 MiB added by this load (186.34 MiB (engine data folder), 176.61 MiB before)" }, "passOrder": [ "default@1", "default@8", "exact" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "totalVectors": 524, "detail": "no index, exact scan by design: gvb_gvbbench_eshoponweb is a vec0 virtual table holding 524 vectors and every search compares all of them (EXPLAIN QUERY PLAN: SCAN gvb_gvbbench_eshoponweb VIRTUAL TABLE INDEX 0:3{___}___ | USE TEMP B-TREE FOR ORDER BY)" }, "afterSearch": { "ready": true, "totalVectors": 524, "detail": "no index, exact scan by design: gvb_gvbbench_eshoponweb is a vec0 virtual table holding 524 vectors and every search compares all of them (EXPLAIN QUERY PLAN: SCAN gvb_gvbbench_eshoponweb VIRTUAL TABLE INDEX 0:3{___}___ | USE TEMP B-TREE FOR ORDER BY)" } }, "durability": "PRAGMA journal_mode=WAL and PRAGMA synchronous=NORMAL (set by this sink when it opens the file): each commit is appended to the -wal file and handed to the operating system, but the WAL is fsynced only when SQLite checkpoints it (measured with strace: 40 single-record commits caused 3 WAL fsyncs), so a crash of the process loses nothing (measured with kill -9 and a reopen: all 50, 300 and 2,000 rows were there), while an operating-system crash or power cut can lose the newest commits since the last checkpoint and leaves the database consistent", "notes": [ "Index after the load: ready, ? of 524 indexed (no index, exact scan by design: gvb_gvbbench_eshoponweb is a vec0 virtual table holding 524 vectors and every search compares all of them (EXPLAIN QUERY PLAN: SCAN gvb_gvbbench_eshoponweb VIRTUAL TABLE INDEX 0:3{___}___ | USE TEMP B-TREE FOR ORDER BY)). Durability: PRAGMA journal_mode=WAL and PRAGMA synchronous=NORMAL (set by this sink when it opens the file): each commit is appended to the -wal file and handed to the operating system, but the WAL is fsynced only when SQLite checkpoints it (measured with strace: 40 single-record commits caused 3 WAL fsyncs), so a crash of the process loses nothing (measured with kill -9 and a reopen: all 50, 300 and 2,000 rows were there), while an operating-system crash or power cut can lose the newest commits since the last checkpoint and leaves the database consistent.", "Outside load when searching began: not known yet (too little sampled history). Load average 4.35 4.08 3.77 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@1 15,293 searches in 30.0 s, default@8 34,273 searches in 30.0 s, exact 15,288 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@1: warm-up 15.0 s and 7,683 searches (0 failed) at 1 searcher; p50 of the last windows 1.912, 1.930, 1.896 ms (windows of at least 2 s and 100 searches); trial of 1,546 searches in 3.0 s at 1 searcher: p50 1.905 ms against the settled p50 1.912 ms, 0% apart (limit 10%); the timed pass p50 1.925 ms, 1% from the settled p50 1.912 ms (limit 10% or 0.1 ms; 0.012 ms); default@8: warm-up 15.0 s and 17,159 searches (0 failed) at 8 searchers; QPS of the last windows 1,148, 1,144, 1,146 (windows of at least 2 s and 100 searches); trial of 3,426 searches in 3.0 s at 8 searchers: 1,141 QPS against the settled 1,146 QPS, 0% apart (limit 10%); the timed pass 1,142 QPS, 0% from the settled 1,146 QPS (limit 10%); exact: warm-up 15.0 s and 7,709 searches (0 failed) at 1 searcher; p50 of the last windows 1.916, 1.903, 1.893 ms (windows of at least 2 s and 100 searches); trial of 1,535 searches in 3.0 s at 1 searcher: p50 1.907 ms against the settled p50 1.903 ms, 0% apart (limit 10%); the timed pass p50 1.906 ms, 0% from the settled p50 1.903 ms (limit 10% or 0.1 ms; 0.003 ms).", "Pass default@1 after rehearsal: 2026-10-06T13:36:06.299Z to 2026-10-06T13:36:26.300Z (20.0 s), 10,198 searches, 0 failed, p50 1.925 ms, mean 1.958 ms, p99 2.342 ms, 509.9 QPS (1000/QPS 1.961 ms).", "Pass default@8 after default@1: 2026-10-06T13:36:44.318Z to 2026-10-06T13:37:04.322Z (20.0 s), 22,843 searches, 0 failed, p50 6.707 ms, mean 6.995 ms, p99 13.767 ms, 1141.9 QPS (1000/QPS 0.876 ms).", "Pass exact after default@8: 2026-10-06T13:37:22.332Z to 2026-10-06T13:38:22.333Z (60.0 s), 30,824 searches, 0 failed, p50 1.906 ms, mean 1.944 ms, p99 2.367 ms, 513.7 QPS (1000/QPS 1.947 ms).", "Passes in the order run: default@1, default@8, exact; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 103,912 sent, 0 failed (warmupErrors). Index after the searches: ready, ? of 524 indexed (no index, exact scan by design: gvb_gvbbench_eshoponweb is a vec0 virtual table holding 524 vectors and every search compares all of them (EXPLAIN QUERY PLAN: SCAN gvb_gvbbench_eshoponweb VIRTUAL TABLE INDEX 0:3{___}___ | USE TEMP B-TREE FOR ORDER BY)).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: in this process (embedded), no network: no network address. No open connection seen (no open TCP connection was found).", "CPU pinning: none: an embedded engine runs inside the benchmark client process, so it shares the client CPUs, CPUs 0-1,4-5.", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@1 3492/3492 MHz, average 3494.1/3493.4, performance, outside load 0.23 before the warm-up, 0.24 during, client CPU 2.067 ms per search; default@8 3492/3492 MHz, average 3492.2/3492.0, performance, outside load 0.22 before the warm-up, 0.23 during, client CPU 3.493 ms per search; exact 3492/3492 MHz, average 3493.3/3493.0, performance, outside load 0.23 before the warm-up, 0.22 during, client CPU 2.039 ms per search." ] }, { "name": "qdrant", "engine": "Qdrant 1.17.0 in container gvb-qdrant (image qdrant/qdrant:v1.17.0@sha256:f1c7272cdac52b38c1a0e89313922d940ba50afd90d593a1605dbbc214e66ffb; cpuset 2-3,6-7 from its creation; 3 search thread(s))", "index": "exact scan: the builder\u0027s sink sends exact=true on every search, so no HNSW graph is used whether or not Qdrant has built one (see the index state)", "hosting": "compose", "searchSettings": { "exact": "true" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.1041073, "rowsPerSecond": 5033.268560417953, "countMatchSeconds": 0.0096455, "indexSeconds": 1.0161006, "indexNote": "status Green, 0 of 524 vectors in HNSW segments (2 segments), waited 1.0 s" }, "search": { "countInTarget": 524, "latencySamples": 22224, "p50Ms": 0.8858, "p95Ms": 1.0155699999999999, "p99Ms": 1.3988390000000004, "qps": { "1": 1111.1778042233607, "8": 3273.8173965216006 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 0, "clientCpuMsPerSearch": { "1": 0.7373, "8": 0.5272 } }, "ram": { "bytes": 94571069, "text": "90.19 MiB (docker stats)" }, "disk": { "bytes": 205603870, "text": "196.08 MiB (collection folder)" }, "passOrder": [ "default@1", "default@8" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 0, "totalVectors": 524, "detail": "status Green, optimizer ok, indexed_vectors_count 0 of 524 points, 2 segments, indexing_threshold_kb 10,000, full_scan_threshold_kb 10,000; no index used, exact scan by design (the builder\u0027s sink sends exact=true on every search); no graph was built; searches counted since the collection was created: unfiltered_hnsw 0, unfiltered_plain 0, unfiltered_exact 0" }, "afterSearch": { "ready": true, "indexedVectors": 0, "totalVectors": 524, "detail": "status Green, optimizer ok, indexed_vectors_count 0 of 524 points, 2 segments, indexing_threshold_kb 10,000, full_scan_threshold_kb 10,000; no index used, exact scan by design (the builder\u0027s sink sends exact=true on every search); no graph was built; searches counted since the collection was created: unfiltered_hnsw 0, unfiltered_plain 597,894, unfiltered_exact 0" } }, "durability": "What was measured (strace -f on the Qdrant server process, 30 single-point REST upserts per setting, every sync call timed against the request that caused it): with wait=true each upsert had exactly one msync(MS_SYNC) of the write-ahead-log segment inside the request, before the reply (30 of 30); with wait=false the 30 upserts were acknowledged within 0.26 s with no sync call, and the first WAL msync (one call covering all 30 records) came 2.7 s after the last reply, followed by the segment-file flushes. So the log is flushed to disk under both settings, but with wait=false the flush comes after the acknowledgement: an operating-system crash or power cut in that gap loses acknowledged writes, while a crash of the Qdrant process alone should not, because the bytes are already in the kernel\u0027s page cache (inferred, not tested). Every upsert here is sent with wait=true in batches of 256 over gRPC; the strace used one point per request, so one flush per batch is inferred, not measured. Segment files are flushed every 5 s; log segments 32 MB, 0 created ahead. /qdrant/config/config.yaml in container gvb-qdrant (the image\u0027s own file; its storage keys are listed) sets: storage.collection.quantization = null, storage.collection.replication_factor = 1, storage.collection.vectors.on_disk = null, storage.collection.write_consistency_factor = 1, storage.hnsw_index.ef_construct = 100, storage.hnsw_index.full_scan_threshold_kb = 10000, storage.hnsw_index.m = 16, storage.hnsw_index.max_indexing_threads = 0, storage.hnsw_index.on_disk = false, storage.hnsw_index.payload_m = null, storage.max_collections = null, storage.node_type = Normal, storage.on_disk_payload = true, storage.optimizers.default_segment_number = 0, storage.optimizers.deleted_threshold = 0.2, storage.optimizers.flush_interval_sec = 5, storage.optimizers.indexing_threshold_kb = 10000, storage.optimizers.max_optimization_threads = null, storage.optimizers.max_segment_size_kb = null, storage.optimizers.vacuum_min_vector_number = 1000, storage.performance.max_search_threads = 0, storage.performance.optimizer_cpu_budget = 0, storage.performance.update_rate_limit = null, storage.shard_transfer_method = null, storage.snapshots_config.snapshots_storage = local, storage.snapshots_path = ./snapshots, storage.storage_path = ./storage, storage.temp_path = null, storage.update_concurrency = null, storage.wal.wal_capacity_mb = 32, storage.wal.wal_segments_ahead = 0; no QDRANT__ environment overrides in the server process. Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started qdrant.compose.yaml with every container created on CPUs 2-3,6-7: gvb-qdrant cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [qdrant.compose.yaml eb57e55e93c1]. This run created the engine from these files.", "Index after the load: ready, 0 of 524 indexed (status Green, optimizer ok, indexed_vectors_count 0 of 524 points, 2 segments, indexing_threshold_kb 10,000, full_scan_threshold_kb 10,000; no index used, exact scan by design (the builder\u0027s sink sends exact=true on every search); no graph was built; searches counted since the collection was created: unfiltered_hnsw 0, unfiltered_plain 0, unfiltered_exact 0). Durability: What was measured (strace -f on the Qdrant server process, 30 single-point REST upserts per setting, every sync call timed against the request that caused it): with wait=true each upsert had exactly one msync(MS_SYNC) of the write-ahead-log segment inside the request, before the reply (30 of 30); with wait=false the 30 upserts were acknowledged within 0.26 s with no sync call, and the first WAL msync (one call covering all 30 records) came 2.7 s after the last reply, followed by the segment-file flushes. So the log is flushed to disk under both settings, but with wait=false the flush comes after the acknowledgement: an operating-system crash or power cut in that gap loses acknowledged writes, while a crash of the Qdrant process alone should not, because the bytes are already in the kernel\u0027s page cache (inferred, not tested). Every upsert here is sent with wait=true in batches of 256 over gRPC; the strace used one point per request, so one flush per batch is inferred, not measured. Segment files are flushed every 5 s; log segments 32 MB, 0 created ahead. /qdrant/config/config.yaml in container gvb-qdrant (the image\u0027s own file; its storage keys are listed) sets: storage.collection.quantization = null, storage.collection.replication_factor = 1, storage.collection.vectors.on_disk = null, storage.collection.write_consistency_factor = 1, storage.hnsw_index.ef_construct = 100, storage.hnsw_index.full_scan_threshold_kb = 10000, storage.hnsw_index.m = 16, storage.hnsw_index.max_indexing_threads = 0, storage.hnsw_index.on_disk = false, storage.hnsw_index.payload_m = null, storage.max_collections = null, storage.node_type = Normal, storage.on_disk_payload = true, storage.optimizers.default_segment_number = 0, storage.optimizers.deleted_threshold = 0.2, storage.optimizers.flush_interval_sec = 5, storage.optimizers.indexing_threshold_kb = 10000, storage.optimizers.max_optimization_threads = null, storage.optimizers.max_segment_size_kb = null, storage.optimizers.vacuum_min_vector_number = 1000, storage.performance.max_search_threads = 0, storage.performance.optimizer_cpu_budget = 0, storage.performance.update_rate_limit = null, storage.shard_transfer_method = null, storage.snapshots_config.snapshots_storage = local, storage.snapshots_path = ./snapshots, storage.storage_path = ./storage, storage.temp_path = null, storage.update_concurrency = null, storage.wal.wal_capacity_mb = 32, storage.wal.wal_segments_ahead = 0; no QDRANT__ environment overrides in the server process. Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked..", "Outside load when searching began: 0.26 CPUs on average over the last 1 s and 0.26 over the last few seconds (limit 0.3), a quiet box. The load average 2.12 3.48 3.63 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@1 33,819 searches in 30.0 s, default@8 98,331 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@1: warm-up 15.0 s and 16,688 searches (0 failed) at 1 searcher; p50 of the last windows 0.882, 0.889, 0.884 ms (windows of at least 2 s and 100 searches); trial of 3,347 searches in 3.0 s at 1 searcher: p50 0.881 ms against the settled p50 0.884 ms, 0% apart (limit 10%); the timed pass p50 0.886 ms, 0% from the settled p50 0.884 ms (limit 10% or 0.1 ms; 0.002 ms); default@8: warm-up 15.0 s and 49,236 searches (0 failed) at 8 searchers; QPS of the last windows 3,277, 3,291, 3,257 (windows of at least 2 s and 100 searches); trial of 9,819 searches in 3.0 s at 8 searchers: 3,271 QPS against the settled 3,277 QPS, 0% apart (limit 10%); the timed pass 3,274 QPS, 0% from the settled 3,277 QPS (limit 10%).", "Pass default@1 after rehearsal: 2026-10-06T13:39:48.157Z to 2026-10-06T13:40:08.157Z (20.0 s), 22,224 searches, 0 failed, p50 0.886 ms, mean 0.898 ms, p99 1.399 ms, 1111.2 QPS (1000/QPS 0.900 ms).", "Pass default@8 after default@1: 2026-10-06T13:40:26.187Z to 2026-10-06T13:40:46.189Z (20.0 s), 65,483 searches, 0 failed, p50 2.357 ms, mean 2.441 ms, p99 4.258 ms, 3273.8 QPS (1000/QPS 0.305 ms).", "Passes in the order run: default@1, default@8; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 211,240 sent, 0 failed (warmupErrors). Index after the searches: ready, 0 of 524 indexed (status Green, optimizer ok, indexed_vectors_count 0 of 524 points, 2 segments, indexing_threshold_kb 10,000, full_scan_threshold_kb 10,000; no index used, exact scan by design (the builder\u0027s sink sends exact=true on every search); no graph was built; searches counted since the collection was created: unfiltered_hnsw 0, unfiltered_plain 597,894, unfiltered_exact 0).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-qdrant (880915fef269) on network gvb-qdrant_default at container-ip:6334, not through docker-proxy localhost:16334; container gvb-qdrant (880915fef269) on network gvb-qdrant_default at container-ip:6333, not through docker-proxy localhost:16333. Open after the passes: container-ip:6334 x1 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-qdrant: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-qdrant: 24 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@1 3492/3492 MHz, average 3492.0/3491.8, performance, outside load 0.19 before the warm-up, 0.18 during, client CPU 0.737 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.0, performance, outside load 0.19 before the warm-up, 0.22 during, client CPU 0.527 ms per search." ] }, { "name": "opensearch", "engine": "OpenSearch 3.9.0 (k-NN plugin, faiss)", "index": "faiss HNSW float32, no compression, m=16, ef_construction=128, cosinesimil; search k=top, ef_search=100; 1 shard, 0 replicas; graph built at any segment size (approximate_threshold=0); force-merged to one segment after the load", "hosting": "compose", "searchSettings": { "approximate_threshold": "0", "ef_construction": "128", "ef_search": "100", "k": "top", "m": "16" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 1.418305, "rowsPerSecond": 369.4550889970775, "countMatchSeconds": 0.1408923, "indexSeconds": 0.8950152, "indexNote": "force-merged to one segment, graph ready after 0.9 s: every segment searched through its HNSW graph, covering 524 of 524 vectors: 1 segment(s), 0 merge(s) running, profiled probe search: ann_search_count 1 (segments tried through a graph), exact_search_count 0 (of those, scanned instead), approximate_threshold 0; force-merged to one segment on purpose, so one graph answers every search" }, "search": { "countInTarget": 524, "latencySamples": 9018, "p50Ms": 2.1568, "p95Ms": 2.49923, "p99Ms": 3.4544729999999997, "qps": { "8": 1346.3120709689424, "1": 450.8554892918197 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 21439, "exactP50Ms": 2.7499, "exactP95Ms": 3.0830599999999992, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 0.5147, "1": 0.473 } }, "ram": { "bytes": 2865816928, "text": "2.67 GiB (docker stats)" }, "disk": { "bytes": 10129139, "text": "9.66 MiB added by this load (12.37 MiB (whole engine data folder), 2.71 MiB before)" }, "passOrder": [ "exact", "default@8", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "every segment searched through its HNSW graph, covering 524 of 524 vectors: 1 segment(s), 0 merge(s) running, profiled probe search: ann_search_count 1 (segments tried through a graph), exact_search_count 0 (of those, scanned instead), approximate_threshold 0; force-merged to one segment on purpose, so one graph answers every search" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "every segment searched through its HNSW graph, covering 524 of 524 vectors: 1 segment(s), 0 merge(s) running, profiled probe search: ann_search_count 1 (segments tried through a graph), exact_search_count 0 (of those, scanned instead), approximate_threshold 0; force-merged to one segment on purpose, so one graph answers every search" } }, "durability": "Every acknowledged bulk request is fsynced to the translog before the answer: index.translog.durability=request, the OpenSearch default, which neither this sink nor opensearch.compose.yaml overrides (OpenSearchReadinessTests reads it back from the live index). By that setting a process crash or power loss loses no acknowledged write; this is read from the setting, not shown by pulling power. One node and no replicas, so a lost disk loses the data.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started opensearch.compose.yaml with every container created on CPUs 2-3,6-7: gvb-opensearch cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [opensearch.compose.yaml 028b44c40129]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (every segment searched through its HNSW graph, covering 524 of 524 vectors: 1 segment(s), 0 merge(s) running, profiled probe search: ann_search_count 1 (segments tried through a graph), exact_search_count 0 (of those, scanned instead), approximate_threshold 0; force-merged to one segment on purpose, so one graph answers every search). Durability: Every acknowledged bulk request is fsynced to the translog before the answer: index.translog.durability=request, the OpenSearch default, which neither this sink nor opensearch.compose.yaml overrides (OpenSearchReadinessTests reads it back from the live index). By that setting a process crash or power loss loses no acknowledged write; this is read from the setting, not shown by pulling power. One node and no replicas, so a lost disk loses the data..", "Outside load when searching began: 0.22 CPUs on average over the last 3 s and 0.22 over the last few seconds (limit 0.3), a quiet box. The load average 4.32 4.02 3.81 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: exact 7,799 searches in 30.0 s, default@8 33,192 searches in 30.0 s, default@1 13,652 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. exact: warm-up 15.0 s and 5,166 searches (0 failed) at 1 searcher; p50 of the last windows 2.797, 2.790, 2.780 ms (windows of at least 2 s and 100 searches); trial of 1,035 searches in 3.0 s at 1 searcher: p50 2.797 ms against the settled p50 2.790 ms, 0% apart (limit 10%); the timed pass p50 2.750 ms, 1% from the settled p50 2.790 ms (limit 10% or 0.1 ms; 0.040 ms); default@8: warm-up 15.0 s and 20,114 searches (0 failed) at 8 searchers; QPS of the last windows 1,338, 1,354, 1,345 (windows of at least 2 s and 100 searches); trial of 4,041 searches in 3.0 s at 8 searchers: 1,345 QPS against the settled 1,345 QPS, 0% apart (limit 10%); the timed pass 1,346 QPS, 0% from the settled 1,345 QPS (limit 10%); default@1: warm-up 15.0 s and 6,839 searches (0 failed) at 1 searcher; p50 of the last windows 2.145, 2.143, 2.144 ms (windows of at least 2 s and 100 searches); trial of 1,379 searches in 3.0 s at 1 searcher: p50 2.141 ms against the settled p50 2.144 ms, 0% apart (limit 10%); the timed pass p50 2.157 ms, 1% from the settled p50 2.144 ms (limit 10% or 0.1 ms; 0.013 ms).", "Pass exact after rehearsal: 2026-10-06T13:43:05.836Z to 2026-10-06T13:44:05.838Z (60.0 s), 21,439 searches, 0 failed, p50 2.750 ms, mean 2.797 ms, p99 3.812 ms, 357.3 QPS (1000/QPS 2.799 ms).", "Pass default@8 after exact: 2026-10-06T13:44:23.868Z to 2026-10-06T13:44:43.873Z (20.0 s), 26,932 searches, 0 failed, p50 5.723 ms, mean 5.939 ms, p99 11.682 ms, 1346.3 QPS (1000/QPS 0.743 ms).", "Pass default@1 after default@8: 2026-10-06T13:45:01.885Z to 2026-10-06T13:45:21.887Z (20.0 s), 9,018 searches, 0 failed, p50 2.157 ms, mean 2.216 ms, p99 3.454 ms, 450.9 QPS (1000/QPS 2.218 ms).", "Passes in the order run: exact, default@8, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 93,217 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (every segment searched through its HNSW graph, covering 524 of 524 vectors: 1 segment(s), 0 merge(s) running, profiled probe search: ann_search_count 1 (segments tried through a graph), exact_search_count 0 (of those, scanned instead), approximate_threshold 0; force-merged to one segment on purpose, so one graph answers every search).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-opensearch (53b5350e8b2f) on network engines_default at container-ip:9200, not through docker-proxy localhost:9201. Open after the passes: container-ip:9200 x8 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-opensearch: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-opensearch: 55 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): exact 3492/3492 MHz, average 3492.1/3492.5, performance, outside load 0.20 before the warm-up, 0.21 during, client CPU 0.474 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.2, performance, outside load 0.21 before the warm-up, 0.20 during, client CPU 0.515 ms per search; default@1 3492/3492 MHz, average 3492.0/3492.6, performance, outside load 0.21 before the warm-up, 0.21 during, client CPU 0.473 ms per search." ] }, { "name": "qdrant-hnsw", "engine": "Qdrant 1.17.0 in container gvb-qdrant (image qdrant/qdrant:v1.17.0@sha256:f1c7272cdac52b38c1a0e89313922d940ba50afd90d593a1605dbbc214e66ffb; cpuset 2-3,6-7 from its creation; 3 search thread(s))", "index": "HNSW m=16 ef_construct=100, hnsw_ef=server default, cosine; indexing_threshold_kb 1 and full_scan_threshold_kb 10 (server defaults are 10,000 each) so a small collection builds and walks its graph", "hosting": "compose", "searchSettings": { "ef_construct": "100", "hnsw_ef": "server", "m": "16" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.0738603, "rowsPerSecond": 7094.474298100603, "countMatchSeconds": 0.0022195, "indexSeconds": 2.007034, "indexNote": "indexing_threshold_kb 1, full_scan_threshold_kb 10; status Green, 524 of 524 vectors in HNSW segments (2 segments), waited 2.0 s" }, "search": { "countInTarget": 524, "latencySamples": 21430, "p50Ms": 0.92, "p95Ms": 1.0469, "p99Ms": 1.4056, "qps": { "1": 1071.4659166691908, "8": 3451.9026013313824 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 66103, "exactP50Ms": 0.8946, "exactP95Ms": 1.0166, "exactRecall": 1, "clientCpuMsPerSearch": { "1": 0.7271, "8": 0.5102 } }, "ram": { "bytes": 38860226, "text": "37.06 MiB (docker stats)" }, "disk": { "bytes": 172067138, "text": "164.1 MiB (collection folder)" }, "passOrder": [ "default@1", "default@8", "exact" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "status Green, optimizer ok, indexed_vectors_count 524 of 524 points, 2 segments, indexing_threshold_kb 1, full_scan_threshold_kb 10; every one of 1 non-empty segments has a complete HNSW graph above its full-scan threshold; no search has run yet, so the walk is proven by the segment settings only; searches counted since the collection was created: unfiltered_hnsw 0, unfiltered_plain 0, unfiltered_exact 0" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "status Green, optimizer ok, indexed_vectors_count 524 of 524 points, 2 segments, indexing_threshold_kb 1, full_scan_threshold_kb 10; every one of 1 non-empty segments has a complete HNSW graph above its full-scan threshold; 306,698 segment searches walked the graph and none scanned; searches counted since the collection was created: unfiltered_hnsw 306,698, unfiltered_plain 0, unfiltered_exact 118,831" } }, "durability": "What was measured (strace -f on the Qdrant server process, 30 single-point REST upserts per setting, every sync call timed against the request that caused it): with wait=true each upsert had exactly one msync(MS_SYNC) of the write-ahead-log segment inside the request, before the reply (30 of 30); with wait=false the 30 upserts were acknowledged within 0.26 s with no sync call, and the first WAL msync (one call covering all 30 records) came 2.7 s after the last reply, followed by the segment-file flushes. So the log is flushed to disk under both settings, but with wait=false the flush comes after the acknowledgement: an operating-system crash or power cut in that gap loses acknowledged writes, while a crash of the Qdrant process alone should not, because the bytes are already in the kernel\u0027s page cache (inferred, not tested). Every upsert here is sent with wait=true in batches of 256 over gRPC; the strace used one point per request, so one flush per batch is inferred, not measured. Segment files are flushed every 5 s; log segments 32 MB, 0 created ahead. /qdrant/config/config.yaml in container gvb-qdrant (the image\u0027s own file; its storage keys are listed) sets: storage.collection.quantization = null, storage.collection.replication_factor = 1, storage.collection.vectors.on_disk = null, storage.collection.write_consistency_factor = 1, storage.hnsw_index.ef_construct = 100, storage.hnsw_index.full_scan_threshold_kb = 10000, storage.hnsw_index.m = 16, storage.hnsw_index.max_indexing_threads = 0, storage.hnsw_index.on_disk = false, storage.hnsw_index.payload_m = null, storage.max_collections = null, storage.node_type = Normal, storage.on_disk_payload = true, storage.optimizers.default_segment_number = 0, storage.optimizers.deleted_threshold = 0.2, storage.optimizers.flush_interval_sec = 5, storage.optimizers.indexing_threshold_kb = 10000, storage.optimizers.max_optimization_threads = null, storage.optimizers.max_segment_size_kb = null, storage.optimizers.vacuum_min_vector_number = 1000, storage.performance.max_search_threads = 0, storage.performance.optimizer_cpu_budget = 0, storage.performance.update_rate_limit = null, storage.shard_transfer_method = null, storage.snapshots_config.snapshots_storage = local, storage.snapshots_path = ./snapshots, storage.storage_path = ./storage, storage.temp_path = null, storage.update_concurrency = null, storage.wal.wal_capacity_mb = 32, storage.wal.wal_segments_ahead = 0; no QDRANT__ environment overrides in the server process. Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started qdrant.compose.yaml with every container created on CPUs 2-3,6-7: gvb-qdrant cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [qdrant.compose.yaml eb57e55e93c1]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (status Green, optimizer ok, indexed_vectors_count 524 of 524 points, 2 segments, indexing_threshold_kb 1, full_scan_threshold_kb 10; every one of 1 non-empty segments has a complete HNSW graph above its full-scan threshold; no search has run yet, so the walk is proven by the segment settings only; searches counted since the collection was created: unfiltered_hnsw 0, unfiltered_plain 0, unfiltered_exact 0). Durability: What was measured (strace -f on the Qdrant server process, 30 single-point REST upserts per setting, every sync call timed against the request that caused it): with wait=true each upsert had exactly one msync(MS_SYNC) of the write-ahead-log segment inside the request, before the reply (30 of 30); with wait=false the 30 upserts were acknowledged within 0.26 s with no sync call, and the first WAL msync (one call covering all 30 records) came 2.7 s after the last reply, followed by the segment-file flushes. So the log is flushed to disk under both settings, but with wait=false the flush comes after the acknowledgement: an operating-system crash or power cut in that gap loses acknowledged writes, while a crash of the Qdrant process alone should not, because the bytes are already in the kernel\u0027s page cache (inferred, not tested). Every upsert here is sent with wait=true in batches of 256 over gRPC; the strace used one point per request, so one flush per batch is inferred, not measured. Segment files are flushed every 5 s; log segments 32 MB, 0 created ahead. /qdrant/config/config.yaml in container gvb-qdrant (the image\u0027s own file; its storage keys are listed) sets: storage.collection.quantization = null, storage.collection.replication_factor = 1, storage.collection.vectors.on_disk = null, storage.collection.write_consistency_factor = 1, storage.hnsw_index.ef_construct = 100, storage.hnsw_index.full_scan_threshold_kb = 10000, storage.hnsw_index.m = 16, storage.hnsw_index.max_indexing_threads = 0, storage.hnsw_index.on_disk = false, storage.hnsw_index.payload_m = null, storage.max_collections = null, storage.node_type = Normal, storage.on_disk_payload = true, storage.optimizers.default_segment_number = 0, storage.optimizers.deleted_threshold = 0.2, storage.optimizers.flush_interval_sec = 5, storage.optimizers.indexing_threshold_kb = 10000, storage.optimizers.max_optimization_threads = null, storage.optimizers.max_segment_size_kb = null, storage.optimizers.vacuum_min_vector_number = 1000, storage.performance.max_search_threads = 0, storage.performance.optimizer_cpu_budget = 0, storage.performance.update_rate_limit = null, storage.shard_transfer_method = null, storage.snapshots_config.snapshots_storage = local, storage.snapshots_path = ./snapshots, storage.storage_path = ./storage, storage.temp_path = null, storage.update_concurrency = null, storage.wal.wal_capacity_mb = 32, storage.wal.wal_segments_ahead = 0; no QDRANT__ environment overrides in the server process. Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked..", "Outside load when searching began: 0.17 CPUs on average over the last 2 s and 0.17 over the last few seconds (limit 0.3), a quiet box. The load average 2.81 3.61 3.70 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@1 32,060 searches in 30.0 s, default@8 103,437 searches in 30.0 s, exact 32,979 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@1: warm-up 15.0 s and 16,078 searches (0 failed) at 1 searcher; p50 of the last windows 0.919, 0.920, 0.923 ms (windows of at least 2 s and 100 searches); trial of 3,227 searches in 3.0 s at 1 searcher: p50 0.921 ms against the settled p50 0.920 ms, 0% apart (limit 10%); the timed pass p50 0.920 ms, 0% from the settled p50 0.920 ms (limit 10% or 0.1 ms; 0.000 ms); default@8: warm-up 15.0 s and 51,188 searches (0 failed) at 8 searchers; QPS of the last windows 3,411, 3,412, 3,425 (windows of at least 2 s and 100 searches); trial of 10,234 searches in 3.0 s at 8 searchers: 3,410 QPS against the settled 3,412 QPS, 0% apart (limit 10%); the timed pass 3,452 QPS, 1% from the settled 3,412 QPS (limit 10%); exact: warm-up 15.0 s and 16,449 searches (0 failed) at 1 searcher; p50 of the last windows 0.899, 0.898, 0.899 ms (windows of at least 2 s and 100 searches); trial of 3,300 searches in 3.0 s at 1 searcher: p50 0.897 ms against the settled p50 0.899 ms, 0% apart (limit 10%); the timed pass p50 0.895 ms, 0% from the settled p50 0.899 ms (limit 10% or 0.1 ms; 0.004 ms).", "Pass default@1 after rehearsal: 2026-10-06T13:47:20.324Z to 2026-10-06T13:47:40.325Z (20.0 s), 21,430 searches, 0 failed, p50 0.920 ms, mean 0.932 ms, p99 1.406 ms, 1071.5 QPS (1000/QPS 0.933 ms).", "Pass default@8 after default@1: 2026-10-06T13:47:58.352Z to 2026-10-06T13:48:18.354Z (20.0 s), 69,044 searches, 0 failed, p50 2.234 ms, mean 2.315 ms, p99 3.999 ms, 3451.9 QPS (1000/QPS 0.290 ms).", "Pass exact after default@8: 2026-10-06T13:48:36.383Z to 2026-10-06T13:49:36.383Z (60.0 s), 66,103 searches, 0 failed, p50 0.895 ms, mean 0.906 ms, p99 1.389 ms, 1101.7 QPS (1000/QPS 0.908 ms).", "Passes in the order run: default@1, default@8, exact; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 268,952 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (status Green, optimizer ok, indexed_vectors_count 524 of 524 points, 2 segments, indexing_threshold_kb 1, full_scan_threshold_kb 10; every one of 1 non-empty segments has a complete HNSW graph above its full-scan threshold; 306,698 segment searches walked the graph and none scanned; searches counted since the collection was created: unfiltered_hnsw 306,698, unfiltered_plain 0, unfiltered_exact 118,831).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-qdrant (69832df26c6f) on network gvb-qdrant_default at container-ip:6334, not through docker-proxy localhost:16334; container gvb-qdrant (69832df26c6f) on network gvb-qdrant_default at container-ip:6333, not through docker-proxy localhost:16333. Open after the passes: container-ip:6334 x1 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-qdrant: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-qdrant: 24 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@1 3492/3492 MHz, average 3492.0/3492.1, performance, outside load 0.20 before the warm-up, 0.17 during, client CPU 0.727 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.0, performance, outside load 0.16 before the warm-up, 0.23 during, client CPU 0.510 ms per search; exact 3492/3492 MHz, average 3492.0/3491.9, performance, outside load 0.22 before the warm-up, 0.17 during, client CPU 0.730 ms per search." ] }, { "name": "redis", "engine": "Redis 8.10.2 (query engine, HASH \u002B vector index)", "index": "HNSW TYPE FLOAT32 M=16 EF_CONSTRUCTION=128, EF_RUNTIME=100 per query, cosine; exact mode = FLAT index built on first exact query", "hosting": "compose", "searchSettings": { "EF_CONSTRUCTION": "128", "EF_RUNTIME": "100", "M": "16" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.0422411, "rowsPerSecond": 12404.979983949283, "countMatchSeconds": 1.021156, "indexSeconds": 0.0240399, "indexNote": "HNSW index finished after 0.0 s of waiting: FT.INFO indexing 0, percent_indexed 1, hash_indexing_failures 0, flat_buffer_size 0, num_docs 524, hashes counted with SCAN under gvb_gvbbench_eshoponweb: 524" }, "search": { "countInTarget": 524, "latencySamples": 50539, "p50Ms": 0.3978, "p95Ms": 0.48900999999999983, "p99Ms": 0.7173620000000003, "qps": { "8": 5120.058543295747, "1": 2526.9181229278793 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 159894, "exactP50Ms": 0.3781, "exactP95Ms": 0.469, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 0.364, "1": 0.5687 } }, "ram": { "bytes": 266548019, "text": "254.2 MiB (docker stats)" }, "disk": { "bytes": 0, "text": "0 B added by this load (94.22 MiB (whole engine data folder), 94.22 MiB before)" }, "passOrder": [ "default@8", "default@1", "exact" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "FT.INFO indexing 0, percent_indexed 1, hash_indexing_failures 0, flat_buffer_size 0, num_docs 524, hashes counted with SCAN under gvb_gvbbench_eshoponweb: 524" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "FT.INFO indexing 0, percent_indexed 1, hash_indexing_failures 0, flat_buffer_size 0, num_docs 524, hashes counted with SCAN under gvb_gvbbench_eshoponweb: 524" } }, "durability": "save \u0022300 1\u0022 and appendonly no (redis.compose.yaml): an RDB snapshot is written every 5 minutes if at least one key changed, and on a clean stop; there is no append-only log, so a crash loses every write since the last snapshot (up to 5 minutes plus the time a snapshot takes; measured with docker kill, which is SIGKILL, right after a 2,000-vector load: none of the 2,000 hashes were there after the restart, and the restart loaded an older snapshot that still held keys of a collection that had been dropped since)", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started redis.compose.yaml with every container created on CPUs 2-3,6-7: gvb-redis cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [redis.compose.yaml 7448634e0d92]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (FT.INFO indexing 0, percent_indexed 1, hash_indexing_failures 0, flat_buffer_size 0, num_docs 524, hashes counted with SCAN under gvb_gvbbench_eshoponweb: 524). Durability: save \u0022300 1\u0022 and appendonly no (redis.compose.yaml): an RDB snapshot is written every 5 minutes if at least one key changed, and on a clean stop; there is no append-only log, so a crash loses every write since the last snapshot (up to 5 minutes plus the time a snapshot takes; measured with docker kill, which is SIGKILL, right after a 2,000-vector load: none of the 2,000 hashes were there after the restart, and the restart loaded an older snapshot that still held keys of a collection that had been dropped since).", "Outside load when searching began: 0.21 CPUs on average over the last 1 s and 0.21 over the last few seconds (limit 0.3), a quiet box. The load average 3.63 3.65 3.70 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 151,068 searches in 30.0 s, default@1 76,411 searches in 30.0 s, exact 77,196 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 76,663 searches (0 failed) at 8 searchers; QPS of the last windows 5,015, 5,094, 5,132 (windows of at least 2 s and 100 searches); trial of 15,140 searches in 3.0 s at 8 searchers: 5,045 QPS against the settled 5,094 QPS, 1% apart (limit 10%); the timed pass 5,120 QPS, 0% from the settled 5,094 QPS (limit 10%); default@1: warm-up 15.0 s and 37,630 searches (0 failed) at 1 searcher; p50 of the last windows 0.393, 0.407, 0.389 ms (windows of at least 2 s and 100 searches); trial of 7,433 searches in 3.0 s at 1 searcher: p50 0.409 ms against the settled p50 0.393 ms, 4% apart (limit 10%); the timed pass p50 0.398 ms, 1% from the settled p50 0.393 ms (limit 10% or 0.1 ms; 0.005 ms); exact: warm-up 15.0 s and 40,226 searches (0 failed) at 1 searcher; p50 of the last windows 0.372, 0.373, 0.374 ms (windows of at least 2 s and 100 searches); trial of 8,019 searches in 3.0 s at 1 searcher: p50 0.375 ms against the settled p50 0.373 ms, 1% apart (limit 10%); the timed pass p50 0.378 ms, 1% from the settled p50 0.373 ms (limit 10% or 0.1 ms; 0.005 ms).", "Pass default@8 after rehearsal: 2026-10-06T13:51:46.664Z to 2026-10-06T13:52:06.665Z (20.0 s), 102,408 searches, 0 failed, p50 1.564 ms, mean 1.561 ms, p99 2.227 ms, 5120.1 QPS (1000/QPS 0.195 ms).", "Pass default@1 after default@8: 2026-10-06T13:52:24.709Z to 2026-10-06T13:52:44.709Z (20.0 s), 50,539 searches, 0 failed, p50 0.398 ms, mean 0.394 ms, p99 0.717 ms, 2526.9 QPS (1000/QPS 0.396 ms).", "Pass exact after default@1: 2026-10-06T13:53:02.775Z to 2026-10-06T13:54:02.775Z (60.0 s), 159,894 searches, 0 failed, p50 0.378 ms, mean 0.374 ms, p99 0.675 ms, 2664.9 QPS (1000/QPS 0.375 ms).", "Passes in the order run: default@8, default@1, exact; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 489,786 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (FT.INFO indexing 0, percent_indexed 1, hash_indexing_failures 0, flat_buffer_size 0, num_docs 524, hashes counted with SCAN under gvb_gvbbench_eshoponweb: 524).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-redis (26d9ff0dafd6) on network gvb-redis_default at container-ip:6379, not through docker-proxy localhost:6379. Open after the passes: container-ip:6379 x1 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-redis: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-redis: 16 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3492.0/3492.0, performance, outside load 0.19 before the warm-up after waiting 3 s, 0.20 during, client CPU 0.364 ms per search; default@1 3492/3492 MHz, average 3492.1/3492.0, performance, outside load 0.20 before the warm-up, 0.20 during, client CPU 0.569 ms per search; exact 3492/3492 MHz, average 3492.1/3492.0, performance, outside load 0.21 before the warm-up, 0.20 during, client CPU 0.565 ms per search." ] }, { "name": "milvus", "engine": "Milvus 2.6.25 (standalone, embedded etcd, local storage, REST API v2)", "index": "HNSW M=16 efConstruction=128, ef=100, metric COSINE, Strong consistency searches; approximate only (no exact mode)", "hosting": "compose", "searchSettings": { "M": "16", "ef": "100", "efConstruction": "128" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.4535577, "rowsPerSecond": 1155.3105591636963, "countMatchSeconds": 1.5965788, "indexSeconds": 138.5606596, "indexNote": "index Finished, type HNSW (COSINE, {\u0022M\u0022:16,\u0022efConstruction\u0022:128}), indexedRows 524 of 524 sealed rows, pendingRows 0; stored rows 524; LoadStateLoaded; query node: 1 sealed segment(s) with the index loaded covering 524 rows against 524 stored rows, 0 sealed without it, 0 rows in growing segments; datacoord: 1 live segment(s), 1 flushed and indexed covering 524 rows, 0 delete-log (L0) segment(s) waiting for a compaction, 3 segment(s) compacted away so far; flush, index build and load took 14.7 s, then compaction: 1 job(s) accepted by Milvus, layout ready and unchanged for 75 s, 123.9 s for the whole step; search ledger started, later state reads count the searches against the query node\u0027s own counters" }, "search": { "countInTarget": 524, "latencySamples": 8108, "p50Ms": 2.3394500000000003, "p95Ms": 2.90236, "p99Ms": 3.9120220000000017, "qps": { "8": 1254.7267841051075, "1": 405.3785108851379 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 0, "clientCpuMsPerSearch": { "8": 0.6384, "1": 0.5686 } }, "ram": { "bytes": 234461593, "text": "223.6 MiB (docker stats)" }, "disk": { "bytes": 11464635, "text": "10.93 MiB added by this load (202.27 MiB (whole engine data folder), 191.34 MiB before)" }, "passOrder": [ "default@8", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "index Finished, type HNSW (COSINE, {\u0022M\u0022:16,\u0022efConstruction\u0022:128}), indexedRows 524 of 524 sealed rows, pendingRows 0; stored rows 524; LoadStateLoaded; query node: 1 sealed segment(s) with the index loaded covering 524 rows against 524 stored rows, 0 sealed without it, 0 rows in growing segments; datacoord: 1 live segment(s), 1 flushed and indexed covering 524 rows, 0 delete-log (L0) segment(s) waiting for a compaction, 3 segment(s) compacted away so far; search ledger since the index finished: the sink sent 0 searches (searchParams.params.ef=100, Strong consistency), the query node counted 0 search requests for this collection, segment searches Sealed \u002B0 and Growing \u002B0, the same sealed segments with the same loaded index builds at both readings, segments compacted away \u002B0" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "index Finished, type HNSW (COSINE, {\u0022M\u0022:16,\u0022efConstruction\u0022:128}), indexedRows 524 of 524 sealed rows, pendingRows 0; stored rows 524; LoadStateLoaded; query node: 1 sealed segment(s) with the index loaded covering 524 rows against 524 stored rows, 0 sealed without it, 0 rows in growing segments; datacoord: 1 live segment(s), 1 flushed and indexed covering 524 rows, 0 delete-log (L0) segment(s) waiting for a compaction, 3 segment(s) compacted away so far; search ledger since the index finished: the sink sent 112786 searches (searchParams.params.ef=100, Strong consistency), the query node counted 112786 search requests for this collection, segment searches Sealed \u002B110583 and Growing \u002B0, the same sealed segments with the same loaded index builds at both readings, segments compacted away \u002B0" } }, "durability": "Writes are not fsynced. Standalone uses the default message queue rocksmq (RocksDB, /var/lib/milvus/rdb_data, mq.type default) and flushed segments go to local-disk object storage (COMMON_STORAGETYPE=local in milvus.compose.yaml). Measured 2026-10-04 with strace over 30 acknowledged single-row upserts and one flush: fdatasync ran only on the embedded etcd files (member/wal, member/snap/db), never on rdb_data or the segment files. A host power loss or kernel crash can lose acknowledged rows that are still in the OS page cache; a Milvus process crash alone should not (inferred, not tested). Metadata in embedded etcd is fdatasynced on every commit.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started milvus.compose.yaml with every container created on CPUs 2-3,6-7: gvb-milvus cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [milvus.compose.yaml eaefecdc3671]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (index Finished, type HNSW (COSINE, {\u0022M\u0022:16,\u0022efConstruction\u0022:128}), indexedRows 524 of 524 sealed rows, pendingRows 0; stored rows 524; LoadStateLoaded; query node: 1 sealed segment(s) with the index loaded covering 524 rows against 524 stored rows, 0 sealed without it, 0 rows in growing segments; datacoord: 1 live segment(s), 1 flushed and indexed covering 524 rows, 0 delete-log (L0) segment(s) waiting for a compaction, 3 segment(s) compacted away so far; search ledger since the index finished: the sink sent 0 searches (searchParams.params.ef=100, Strong consistency), the query node counted 0 search requests for this collection, segment searches Sealed \u002B0 and Growing \u002B0, the same sealed segments with the same loaded index builds at both readings, segments compacted away \u002B0). Durability: Writes are not fsynced. Standalone uses the default message queue rocksmq (RocksDB, /var/lib/milvus/rdb_data, mq.type default) and flushed segments go to local-disk object storage (COMMON_STORAGETYPE=local in milvus.compose.yaml). Measured 2026-10-04 with strace over 30 acknowledged single-row upserts and one flush: fdatasync ran only on the embedded etcd files (member/wal, member/snap/db), never on rdb_data or the segment files. A host power loss or kernel crash can lose acknowledged rows that are still in the OS page cache; a Milvus process crash alone should not (inferred, not tested). Metadata in embedded etcd is fdatasynced on every commit..", "Outside load when searching began: 0.20 CPUs on average over the last 60 s and 0.21 over the last few seconds (limit 0.3), a quiet box. The load average 0.56 2.12 3.05 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 37,601 searches in 30.0 s, default@1 12,125 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 18,832 searches (0 failed) at 8 searchers; QPS of the last windows 1,272, 1,259, 1,258 (windows of at least 2 s and 100 searches); trial of 3,775 searches in 3.0 s at 8 searchers: 1,257 QPS against the settled 1,259 QPS, 0% apart (limit 10%); the timed pass 1,255 QPS, 0% from the settled 1,259 QPS (limit 10%); default@1: warm-up 15.0 s and 6,034 searches (0 failed) at 1 searcher; p50 of the last windows 2.348, 2.352, 2.367 ms (windows of at least 2 s and 100 searches); trial of 1,212 searches in 3.0 s at 1 searcher: p50 2.349 ms against the settled p50 2.352 ms, 0% apart (limit 10%); the timed pass p50 2.339 ms, 1% from the settled p50 2.352 ms (limit 10% or 0.1 ms; 0.013 ms).", "Pass default@8 after rehearsal: 2026-10-06T13:57:51.754Z to 2026-10-06T13:58:11.758Z (20.0 s), 25,099 searches, 0 failed, p50 5.834 ms, mean 6.373 ms, p99 14.120 ms, 1254.7 QPS (1000/QPS 0.797 ms).", "Pass default@1 after default@8: 2026-10-06T13:58:29.776Z to 2026-10-06T13:58:49.777Z (20.0 s), 8,108 searches, 0 failed, p50 2.339 ms, mean 2.464 ms, p99 3.912 ms, 405.4 QPS (1000/QPS 2.467 ms).", "Passes in the order run: default@8, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 79,579 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (index Finished, type HNSW (COSINE, {\u0022M\u0022:16,\u0022efConstruction\u0022:128}), indexedRows 524 of 524 sealed rows, pendingRows 0; stored rows 524; LoadStateLoaded; query node: 1 sealed segment(s) with the index loaded covering 524 rows against 524 stored rows, 0 sealed without it, 0 rows in growing segments; datacoord: 1 live segment(s), 1 flushed and indexed covering 524 rows, 0 delete-log (L0) segment(s) waiting for a compaction, 3 segment(s) compacted away so far; search ledger since the index finished: the sink sent 112786 searches (searchParams.params.ef=100, Strong consistency), the query node counted 112786 search requests for this collection, segment searches Sealed \u002B110583 and Growing \u002B0, the same sealed segments with the same loaded index builds at both readings, segments compacted away \u002B0).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-milvus (b9f11218b6aa) on network gvb-milvus_default at container-ip:19530, not through docker-proxy localhost:19530; container gvb-milvus (b9f11218b6aa) on network gvb-milvus_default at container-ip:9091, not through docker-proxy localhost:9091. Open after the passes: container-ip:19530 x8 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-milvus: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-milvus: 24 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3492.0/3492.3, performance, outside load 0.23 before the warm-up, 0.25 during, client CPU 0.638 ms per search; default@1 3492/3492 MHz, average 3492.0/3492.2, performance, outside load 0.23 before the warm-up, 0.19 during, client CPU 0.569 ms per search." ] }, { "name": "mariadb", "engine": "MariaDB 11.8.9 (InnoDB \u002B VECTOR INDEX)", "index": "VECTOR INDEX (HNSW variant) DISTANCE=cosine, M=16 (no ef_construction setting exists), mhnsw_ef_search=100 per statement (the ef 100 most engines here use, so the search effort matches; MariaDB\u0027s own default is 20; recall@10 at ef 100 falls as the set grows (random 1024-dimension vectors, measured 2026-10-04: 0.99 at 524, 0.89 to 0.92 at 2,000; an earlier run gave about 0.09 at 100,000)), mhnsw_max_cache_size 4G; exact mode = IGNORE INDEX full scan", "hosting": "compose", "searchSettings": { "DISTANCE": "cosine", "M": "16", "mhnsw_ef_search": "100" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.5188038, "rowsPerSecond": 1010.0157323442888, "countMatchSeconds": 0.0050738, "indexSeconds": 0.5294481, "indexNote": "VECTOR INDEX is maintained inside every INSERT, nothing to build; confirmed after 2 poll(s) in 0.5 s (poll 1 of 2 said: not ready: a search at mhnsw_ef_search 100 returned only 393 of 524 requested rows through the index (EXPLAIN key \u0027vec_idx\u0027) (information_schema lists vec_idx as INDEX_TYPE VECTOR, EXPLAIN of the default search uses key vec_idx, the server applied mhnsw_ef_search 100 (asked for 100, read back from the server), a search at that effort returned 393 of 524 requested rows through the index; MariaDB has no per-index row count (the graph is a hidden InnoDB table), so indexed vectors are not reported)): information_schema lists vec_idx as INDEX_TYPE VECTOR, EXPLAIN of the default search uses key vec_idx, the server applied mhnsw_ef_search 100 (asked for 100, read back from the server), a search at that effort returned 521 of 524 requested rows through the index; MariaDB has no per-index row count (the graph is a hidden InnoDB table), so indexed vectors are not reported" }, "search": { "countInTarget": 524, "latencySamples": 30928, "p50Ms": 0.6364, "p95Ms": 0.7188, "p99Ms": 0.8624, "qps": { "8": 5911.461450602725, "1": 1546.3727529120936 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 35976, "exactP50Ms": 1.6391, "exactP95Ms": 1.804025, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 0.3757, "1": 0.4858 } }, "ram": { "bytes": 167247872, "text": "159.5 MiB (docker stats)" }, "disk": { "bytes": 23077570, "text": "22.01 MiB added by this load (182.79 MiB (whole engine data folder), 160.78 MiB before)" }, "passOrder": [ "default@8", "default@1", "exact" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "totalVectors": 524, "detail": "information_schema lists vec_idx as INDEX_TYPE VECTOR, EXPLAIN of the default search uses key vec_idx, the server applied mhnsw_ef_search 100 (asked for 100, read back from the server), a search at that effort returned 521 of 524 requested rows through the index; MariaDB has no per-index row count (the graph is a hidden InnoDB table), so indexed vectors are not reported" }, "afterSearch": { "ready": true, "totalVectors": 524, "detail": "information_schema lists vec_idx as INDEX_TYPE VECTOR, EXPLAIN of the default search uses key vec_idx, the server applied mhnsw_ef_search 100 (asked for 100, read back from the server), a search at that effort returned 521 of 524 requested rows through the index; MariaDB has no per-index row count (the graph is a hidden InnoDB table), so indexed vectors are not reported" } }, "durability": "innodb_flush_log_at_trx_commit=2 (set in mariadb-bench.compose.yaml for the benchmark\u0027s own container gvbbench-mariadb, and in mariadb.compose.yaml for the daily gvb-mariadb, which has the same server settings): the InnoDB redo log is written to the operating system at every commit but fsynced about once a second, so a crash of the mariadbd process loses nothing, while an operating-system crash or power cut can lose the last second of commits; innodb_doublewrite is on (default), the binary log is off, and the vector graph is an InnoDB table under the same log (settings read from the running server with SHOW VARIABLES; the crash behaviour is InnoDB\u0027s documented behaviour for this setting and was not tested on either container)", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started mariadb-bench.compose.yaml with every container created on CPUs 2-3,6-7: gvbbench-mariadb cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [mariadb-bench.compose.yaml e9a2147e6f91]. This run created the engine from these files.", "Index after the load: ready, ? of 524 indexed (information_schema lists vec_idx as INDEX_TYPE VECTOR, EXPLAIN of the default search uses key vec_idx, the server applied mhnsw_ef_search 100 (asked for 100, read back from the server), a search at that effort returned 521 of 524 requested rows through the index; MariaDB has no per-index row count (the graph is a hidden InnoDB table), so indexed vectors are not reported). Durability: innodb_flush_log_at_trx_commit=2 (set in mariadb-bench.compose.yaml for the benchmark\u0027s own container gvbbench-mariadb, and in mariadb.compose.yaml for the daily gvb-mariadb, which has the same server settings): the InnoDB redo log is written to the operating system at every commit but fsynced about once a second, so a crash of the mariadbd process loses nothing, while an operating-system crash or power cut can lose the last second of commits; innodb_doublewrite is on (default), the binary log is off, and the vector graph is an InnoDB table under the same log (settings read from the running server with SHOW VARIABLES; the crash behaviour is InnoDB\u0027s documented behaviour for this setting and was not tested on either container).", "WARNING: processes outside the benchmark used 0.39 CPUs on average over the last 1 s and 0.39 over the last few seconds when searching began (limit 0.3): the box was busy as the untimed rehearsal started. Each timed pass still waits for a quiet box before its warm-up, and one that ran busy is flagged \u0027busy box\u0027 on its own. The load average 2.54 2.75 3.17 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 178,397 searches in 30.0 s, default@1 47,035 searches in 30.0 s, exact 17,961 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 88,349 searches (0 failed) at 8 searchers; QPS of the last windows 5,840, 5,941, 5,954 (windows of at least 2 s and 100 searches); trial of 17,798 searches in 3.0 s at 8 searchers: 5,931 QPS against the settled 5,941 QPS, 0% apart (limit 10%); the timed pass 5,911 QPS, 0% from the settled 5,941 QPS (limit 10%); default@1: warm-up 15.0 s and 23,186 searches (0 failed) at 1 searcher; p50 of the last windows 0.625, 0.641, 0.641 ms (windows of at least 2 s and 100 searches); trial of 4,600 searches in 3.0 s at 1 searcher: p50 0.639 ms against the settled p50 0.641 ms, 0% apart (limit 10%); the timed pass p50 0.636 ms, 1% from the settled p50 0.641 ms (limit 10% or 0.1 ms; 0.004 ms); exact: warm-up 15.0 s and 8,963 searches (0 failed) at 1 searcher; p50 of the last windows 1.635, 1.643, 1.653 ms (windows of at least 2 s and 100 searches); trial of 1,802 searches in 3.0 s at 1 searcher: p50 1.631 ms against the settled p50 1.643 ms, 1% apart (limit 10%); the timed pass p50 1.639 ms, 0% from the settled p50 1.643 ms (limit 10% or 0.1 ms; 0.004 ms).", "Pass default@8 after rehearsal: 2026-10-06T14:00:53.785Z to 2026-10-06T14:01:13.786Z (20.0 s), 118,236 searches, 0 failed, p50 1.305 ms, mean 1.352 ms, p99 2.471 ms, 5911.5 QPS (1000/QPS 0.169 ms).", "Pass default@1 after default@8: 2026-10-06T14:01:31.833Z to 2026-10-06T14:01:51.834Z (20.0 s), 30,928 searches, 0 failed, p50 0.636 ms, mean 0.645 ms, p99 0.862 ms, 1546.4 QPS (1000/QPS 0.647 ms).", "Pass exact after default@1: 2026-10-06T14:02:09.870Z to 2026-10-06T14:03:09.872Z (60.0 s), 35,976 searches, 0 failed, p50 1.639 ms, mean 1.666 ms, p99 2.116 ms, 599.6 QPS (1000/QPS 1.668 ms).", "Passes in the order run: default@8, default@1, exact; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 388,091 sent, 0 failed (warmupErrors). Index after the searches: ready, ? of 524 indexed (information_schema lists vec_idx as INDEX_TYPE VECTOR, EXPLAIN of the default search uses key vec_idx, the server applied mhnsw_ef_search 100 (asked for 100, read back from the server), a search at that effort returned 521 of 524 requested rows through the index; MariaDB has no per-index row count (the graph is a hidden InnoDB table), so indexed vectors are not reported).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvbbench-mariadb (52cb67e69c35) on network gvbbench-mariadb_default at container-ip:3306, not through docker-proxy localhost:13306. Open after the passes: container-ip:3306 x8 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvbbench-mariadb: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvbbench-mariadb: 13 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3492.0/3492.0, performance, outside load 0.20 before the warm-up, 0.16 during, client CPU 0.376 ms per search; default@1 3492/3492 MHz, average 3492.7/3492.2, performance, outside load 0.17 before the warm-up, 0.19 during, client CPU 0.486 ms per search; exact 3492/3492 MHz, average 3492.9/3492.4, performance, outside load 0.19 before the warm-up, 0.19 during, client CPU 0.499 ms per search." ] }, { "name": "weaviate", "engine": "Weaviate 1.39.8 (single node, REST \u002B GraphQL)", "index": "HNSW maxConnections(M)=16 efConstruction=128, ef=-1 (dynamic: limit x 8 clamped 100..500), cosine, no quantization; approximate only (no exact mode)", "hosting": "compose", "searchSettings": { "ef": "-1", "efConstruction": "128" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.5216995, "rowsPerSecond": 1004.409626614555, "countMatchSeconds": 0.0055263, "indexSeconds": 0.0237743, "indexNote": "HNSW is updated inside every batch, nothing to build; confirmed in 0.0 s: shard 9u0EFW2dYzBR vectorIndexingStatus READY, vectorQueueLength 0, node status objectCount 0 (refreshed only when the memtable is flushed, so it lags a minute and does not decide readiness), schema shard 9u0EFW2dYzBR status READY, Aggregate count 524, Aggregate nearVector with objectLimit 524 reached 524, Weaviate reports no count of vectors in the HNSW index, so indexed vectors are not reported" }, "search": { "countInTarget": 524, "latencySamples": 3313, "p50Ms": 5.3612, "p95Ms": 10.7284, "p99Ms": 11.817456, "qps": { "1": 165.64498758267575, "8": 304.9206911223285 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 0, "clientCpuMsPerSearch": { "1": 0.5722, "8": 0.6173 } }, "ram": { "bytes": 184444518, "text": "175.9 MiB (docker stats)" }, "disk": { "bytes": 3100538, "text": "2.96 MiB added by this load (5.02 MiB (whole engine data folder), 2.06 MiB before)" }, "passOrder": [ "default@1", "default@8" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "totalVectors": 524, "detail": "shard 9u0EFW2dYzBR vectorIndexingStatus READY, vectorQueueLength 0, node status objectCount 0 (refreshed only when the memtable is flushed, so it lags a minute and does not decide readiness), schema shard 9u0EFW2dYzBR status READY, Aggregate count 524, Aggregate nearVector with objectLimit 524 reached 524, Weaviate reports no count of vectors in the HNSW index, so indexed vectors are not reported" }, "afterSearch": { "ready": true, "totalVectors": 524, "detail": "shard 9u0EFW2dYzBR vectorIndexingStatus READY, vectorQueueLength 0, node status objectCount 524 (refreshed only when the memtable is flushed, so it lags a minute and does not decide readiness), schema shard 9u0EFW2dYzBR status READY, Aggregate count 524, Aggregate nearVector with objectLimit 524 reached 524, Weaviate reports no count of vectors in the HNSW index, so indexed vectors are not reported" } }, "durability": "Weaviate 1.39.8 defaults, weaviate.compose.yaml sets no persistence variable: every object is appended to the LSM write-ahead log with a plain write and no fsync, and the log is fsynced only when its memtable is flushed, 60 seconds after the last write (PERSISTENCE_MEMTABLES_FLUSH_IDLE_AFTER_SECONDS default; measured with strace: 2,080 writes into the objects log during a 2,000-object load, the first fsync 60 s after the last write), so a power cut can lose the last minute of writes; the HNSW commit log is buffered inside the process (77 writes for 2,000 vectors), so a killed process loses the newest vectors from the vector index while their objects survive (measured with docker kill, which is SIGKILL, one second after the load and a restart, two runs each: 0 to 1 of 50, 264 to 267 of 300 and 1,992 to 1,997 of 2,000 stored objects were still reachable through a vector search)", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started weaviate.compose.yaml with every container created on CPUs 2-3,6-7: gvb-weaviate cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [weaviate.compose.yaml d96b7399662a]. This run created the engine from these files.", "Index after the load: ready, ? of 524 indexed (shard 9u0EFW2dYzBR vectorIndexingStatus READY, vectorQueueLength 0, node status objectCount 0 (refreshed only when the memtable is flushed, so it lags a minute and does not decide readiness), schema shard 9u0EFW2dYzBR status READY, Aggregate count 524, Aggregate nearVector with objectLimit 524 reached 524, Weaviate reports no count of vectors in the HNSW index, so indexed vectors are not reported). Durability: Weaviate 1.39.8 defaults, weaviate.compose.yaml sets no persistence variable: every object is appended to the LSM write-ahead log with a plain write and no fsync, and the log is fsynced only when its memtable is flushed, 60 seconds after the last write (PERSISTENCE_MEMTABLES_FLUSH_IDLE_AFTER_SECONDS default; measured with strace: 2,080 writes into the objects log during a 2,000-object load, the first fsync 60 s after the last write), so a power cut can lose the last minute of writes; the HNSW commit log is buffered inside the process (77 writes for 2,000 vectors), so a killed process loses the newest vectors from the vector index while their objects survive (measured with docker kill, which is SIGKILL, one second after the load and a restart, two runs each: 0 to 1 of 50, 264 to 267 of 300 and 1,992 to 1,997 of 2,000 stored objects were still reachable through a vector search).", "Outside load when searching began: 0.18 CPUs on average over the last 1 s and 0.18 over the last few seconds (limit 0.3), a quiet box. The load average 1.35 2.75 3.14 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@1 5,033 searches in 30.0 s, default@8 9,259 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@1: warm-up 15.0 s and 2,501 searches (0 failed) at 1 searcher; p50 of the last windows 5.348, 5.364, 5.353 ms (windows of at least 2 s and 100 searches); trial of 499 searches in 3.0 s at 1 searcher: p50 5.360 ms against the settled p50 5.353 ms, 0% apart (limit 10%); the timed pass p50 5.361 ms, 0% from the settled p50 5.353 ms (limit 10% or 0.1 ms; 0.008 ms); default@8: warm-up 15.0 s and 4,572 searches (0 failed) at 8 searchers; QPS of the last windows 304, 304, 306 (windows of at least 2 s and 100 searches); trial of 923 searches in 3.0 s at 8 searchers: 306 QPS against the settled 304 QPS, 0% apart (limit 10%); the timed pass 305 QPS, 0% from the settled 304 QPS (limit 10%).", "Pass default@1 after rehearsal: 2026-10-06T14:04:51.911Z to 2026-10-06T14:05:11.911Z (20.0 s), 3,313 searches, 0 failed, p50 5.361 ms, mean 6.034 ms, p99 11.817 ms, 165.6 QPS (1000/QPS 6.037 ms).", "Pass default@8 after default@1: 2026-10-06T14:05:29.950Z to 2026-10-06T14:05:49.968Z (20.0 s), 6,104 searches, 0 failed, p50 24.783 ms, mean 26.229 ms, p99 54.394 ms, 304.9 QPS (1000/QPS 3.280 ms).", "Passes in the order run: default@1, default@8; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 22,787 sent, 0 failed (warmupErrors). Index after the searches: ready, ? of 524 indexed (shard 9u0EFW2dYzBR vectorIndexingStatus READY, vectorQueueLength 0, node status objectCount 524 (refreshed only when the memtable is flushed, so it lags a minute and does not decide readiness), schema shard 9u0EFW2dYzBR status READY, Aggregate count 524, Aggregate nearVector with objectLimit 524 reached 524, Weaviate reports no count of vectors in the HNSW index, so indexed vectors are not reported).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-weaviate (41c8b7799a14) on network gvb-weaviate_default at container-ip:8080, not through docker-proxy localhost:8085. Open after the passes: container-ip:8080 x8 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-weaviate: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-weaviate: 9 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@1 3492/3492 MHz, average 3492.1/3492.6, performance, outside load 0.20 before the warm-up, 0.20 during, client CPU 0.572 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.4, performance, outside load 0.21 before the warm-up, 0.23 during, client CPU 0.617 ms per search." ] }, { "name": "mongodb", "engine": "MongoDB 8.0.32 Atlas Local (mongod \u002B mongot Vector Search)", "index": "vectorSearch index, HNSW maxEdges=16 numEdgeCandidates=128, float32 binData, cosine, numCandidates=20x hits (min 100); exact mode = $vectorSearch exact:true", "hosting": "compose", "searchSettings": { "maxEdges": "16", "numCandidates": "20x", "numEdgeCandidates": "128" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.1478191, "rowsPerSecond": 3544.8734297529886, "countMatchSeconds": 1.7360735, "indexSeconds": 10.2544566, "indexNote": "index status READY, queryable True; mongot holds 524 of 524 documents in 4 segment(s) (269 \u002B 251 \u002B 2 \u002B 2 documents), 4 searched through the HNSW graph (Approximate); segment layout unchanged for 10.2 s over 11 readings; wait took 10.3 s" }, "search": { "countInTarget": 524, "latencySamples": 13821, "p50Ms": 1.4136, "p95Ms": 1.6564, "p99Ms": 1.9464599999999992, "qps": { "1": 691.0434489081044, "8": 2066.043956908663 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 47151, "exactP50Ms": 1.2423, "exactP95Ms": 1.45965, "exactRecall": 1, "clientCpuMsPerSearch": { "1": 0.3118, "8": 0.2875 } }, "ram": { "bytes": 925368320, "text": "882.5 MiB (docker stats)" }, "disk": { "bytes": 175496, "text": "171.38 KiB added by this load (1.37 GiB (whole engine data folder), 1.37 GiB before)" }, "passOrder": [ "default@1", "exact", "default@8" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "index status READY, queryable True; mongot holds 524 of 524 documents in 4 segment(s) (269 \u002B 251 \u002B 2 \u002B 2 documents), 4 searched through the HNSW graph (Approximate)" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "index status READY, queryable True; mongot holds 524 of 524 documents in 4 segment(s) (269 \u002B 251 \u002B 2 \u002B 2 documents), 4 searched through the HNSW graph (Approximate)" } }, "durability": "Acknowledged writes are journaled before the acknowledgement. The sink sets no write concern, so the server default applies: getDefaultRWConcern gives w majority and the one-member replica set (--replSet gvbmongo in the image) has writeConcernMajorityJournalDefault true, with WiredTiger journaling on (journalCommitInterval 100 ms is only the interval for unacknowledged work). Measured 2026-10-04: 30 acknowledged single-document inserts raised WiredTiger \u0027log sync operations\u0027 by 30 and strace showed fdatasync on /data/db/journal/WiredTigerLog files. A crash loses no acknowledged write. mongot\u0027s search index is not part of that promise: it follows the collection asynchronously and is rebuilt from it.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started mongodb.compose.yaml with every container created on CPUs 2-3,6-7: gvb-mongodb cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [mongodb.compose.yaml 4ee6f6411f1f]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (index status READY, queryable True; mongot holds 524 of 524 documents in 4 segment(s) (269 \u002B 251 \u002B 2 \u002B 2 documents), 4 searched through the HNSW graph (Approximate)). Durability: Acknowledged writes are journaled before the acknowledgement. The sink sets no write concern, so the server default applies: getDefaultRWConcern gives w majority and the one-member replica set (--replSet gvbmongo in the image) has writeConcernMajorityJournalDefault true, with WiredTiger journaling on (journalCommitInterval 100 ms is only the interval for unacknowledged work). Measured 2026-10-04: 30 acknowledged single-document inserts raised WiredTiger \u0027log sync operations\u0027 by 30 and strace showed fdatasync on /data/db/journal/WiredTigerLog files. A crash loses no acknowledged write. mongot\u0027s search index is not part of that promise: it follows the collection asynchronously and is rebuilt from it..", "Outside load when searching began: 0.23 CPUs on average over the last 14 s and 0.24 over the last few seconds (limit 0.3), a quiet box. The load average 2.48 2.68 3.05 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@1 17,315 searches in 30.0 s, exact 22,690 searches in 30.0 s, default@8 60,007 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@1: warm-up 15.0 s and 10,336 searches (0 failed) at 1 searcher; p50 of the last windows 1.408, 1.414, 1.428 ms (windows of at least 2 s and 100 searches); trial of 2,054 searches in 3.0 s at 1 searcher: p50 1.421 ms against the settled p50 1.414 ms, 0% apart (limit 10%); the timed pass p50 1.414 ms, 0% from the settled p50 1.414 ms (limit 10% or 0.1 ms; 0.000 ms); exact: warm-up 15.0 s and 11,731 searches (0 failed) at 1 searcher; p50 of the last windows 1.257, 1.243, 1.241 ms (windows of at least 2 s and 100 searches); trial of 2,356 searches in 3.0 s at 1 searcher: p50 1.248 ms against the settled p50 1.243 ms, 0% apart (limit 10%); the timed pass p50 1.242 ms, 0% from the settled p50 1.243 ms (limit 10% or 0.1 ms; 0.001 ms); default@8: warm-up 15.0 s and 30,502 searches (0 failed) at 8 searchers; QPS of the last windows 2,085, 2,092, 1,842 (windows of at least 2 s and 100 searches) (its last 3 windows differed by more than 5%); trial of 6,110 searches in 3.0 s at 8 searchers: 2,034 QPS against the settled 2,085 QPS, 2% apart (limit 10%), so the warm-up was EXTENDED once (30.0 s and 61,828 searches (0 failed) at 8 searchers; QPS of its older 6 windows 2,058 (windows 2,062, 2,048, 2,054, 2,056, 2,060, 2,070) and of its newer 6 2,055 (windows 2,068, 2,054, 2,052, 2,059, 2,036, 2,064), 0.1% apart (limit 5%) (windows of at least 2 s and 100 searches)); second trial of 6,209 searches in 3.0 s at 8 searchers: 2,068 QPS against the settled 2,055 QPS, 1% apart (limit 10%), outside its newer windows\u0027 range 2,036 to 2,068 QPS; settled after the extension; the timed pass 2,066 QPS, 1% from the settled 2,055 QPS (limit 10%).", "Pass default@1 after rehearsal: 2026-10-06T14:08:01.467Z to 2026-10-06T14:08:21.468Z (20.0 s), 13,821 searches, 0 failed, p50 1.414 ms, mean 1.445 ms, p99 1.946 ms, 691.0 QPS (1000/QPS 1.447 ms).", "Pass exact after default@1: 2026-10-06T14:08:40.485Z to 2026-10-06T14:09:40.485Z (60.0 s), 47,151 searches, 0 failed, p50 1.242 ms, mean 1.270 ms, p99 1.791 ms, 785.8 QPS (1000/QPS 1.273 ms).", "Pass default@8 after exact: 2026-10-06T14:10:51.547Z to 2026-10-06T14:11:11.550Z (20.0 s), 41,327 searches, 0 failed, p50 3.746 ms, mean 3.869 ms, p99 6.907 ms, 2066.0 QPS (1000/QPS 0.484 ms).", "Passes in the order run: default@1, exact, default@8; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 262,254 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (index status READY, queryable True; mongot holds 524 of 524 documents in 4 segment(s) (269 \u002B 251 \u002B 2 \u002B 2 documents), 4 searched through the HNSW graph (Approximate)).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-mongodb (441920b747a8) on network gvb-mongodb_default at container-ip:27017, not through docker-proxy localhost:27017. Open after the passes: container-ip:27017 x10 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-mongodb: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-mongodb: 211 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@1 3492/3492 MHz, average 3492.2/3492.2, performance, outside load 0.26 before the warm-up, 0.29 during, client CPU 0.312 ms per search; exact 3492/3492 MHz, average 3492.2/3491.2, performance, outside load 0.27 before the warm-up after waiting 1 s, 0.31 during, BUSY BOX, client CPU 0.304 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.1, performance, outside load 0.29 before the warm-up after waiting 5 s, 0.23 during, client CPU 0.288 ms per search.", "WARNING: busy box during exact: processes outside the benchmark used 0.31 CPUs on average during the pass, above the limit of 0.3." ] }, { "name": "chroma", "engine": "Chroma 1.4.4 (single node, REST API v2)", "index": "HNSW M=16 ef_construction=128, ef_search=100 (Chroma default), cosine; approximate only (no exact mode)", "hosting": "compose", "searchSettings": { "M": "16", "ef_construction": "128", "ef_search": "100" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.7847452, "rowsPerSecond": 667.7326602316267, "countMatchSeconds": 0.0267332, "indexSeconds": 0.0180353, "indexNote": "Chroma has no index build to wait for and exposes no index status; checked in 0.0 s: Chroma exposes no index status (indexing_status endpoint: HTTP 500 {\u0022error\u0022:\u0022InternalError\u0022,\u0022message\u0022:\u0022Method scout_logs is not implemented\u0022}), so indexed vectors are not reported and ready only means the two counts agree to within 1 percent: count endpoint 524, a vector search asking for 524 results returned 524" }, "search": { "countInTarget": 524, "latencySamples": 9379, "p50Ms": 2.1054, "p95Ms": 2.380400000000001, "p99Ms": 2.653222, "qps": { "8": 777.0508315446011, "1": 468.9144211182977 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 0, "clientCpuMsPerSearch": { "8": 0.4804, "1": 0.4803 } }, "ram": { "bytes": 83382763, "text": "79.52 MiB (docker stats)" }, "disk": { "bytes": 424100, "text": "414.16 KiB added by this load (2.01 GiB (whole engine data folder), 2.01 GiB before)" }, "passOrder": [ "default@8", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "totalVectors": 524, "detail": "Chroma exposes no index status (indexing_status endpoint: HTTP 500 {\u0022error\u0022:\u0022InternalError\u0022,\u0022message\u0022:\u0022Method scout_logs is not implemented\u0022}), so indexed vectors are not reported and ready only means the two counts agree to within 1 percent: count endpoint 524, a vector search asking for 524 results returned 524" }, "afterSearch": { "ready": true, "totalVectors": 524, "detail": "Chroma exposes no index status (indexing_status endpoint: HTTP 500 {\u0022error\u0022:\u0022InternalError\u0022,\u0022message\u0022:\u0022Method scout_logs is not implemented\u0022}), so indexed vectors are not reported and ready only means the two counts agree to within 1 percent: count endpoint 524, a vector search asking for 524 results returned 524" } }, "durability": "SQLite rollback journal with its default synchronous=FULL under the data directory (chroma.compose.yaml sets IS_PERSISTENT=1 and PERSIST_DIRECTORY=/data and no sync setting): every write is committed to chroma.sqlite3 with fsync of the journal, the directory and the database file before the call returns (strace: 15 database fsyncs and 38 journal fsyncs for one create, four 500-vector upserts and one delete), and the HNSW files are written every sync_threshold=1000 vectors and rebuilt from the SQLite log after a crash; measured with kill -9 and a restart: all 50, 300 and 2,000 vectors were still stored and searchable", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started chroma.compose.yaml with every container created on CPUs 2-3,6-7: gvb-chroma cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [chroma.compose.yaml f7ed9d3c8ff7]. This run created the engine from these files.", "Index after the load: ready, ? of 524 indexed (Chroma exposes no index status (indexing_status endpoint: HTTP 500 {\u0022error\u0022:\u0022InternalError\u0022,\u0022message\u0022:\u0022Method scout_logs is not implemented\u0022}), so indexed vectors are not reported and ready only means the two counts agree to within 1 percent: count endpoint 524, a vector search asking for 524 results returned 524). Durability: SQLite rollback journal with its default synchronous=FULL under the data directory (chroma.compose.yaml sets IS_PERSISTENT=1 and PERSIST_DIRECTORY=/data and no sync setting): every write is committed to chroma.sqlite3 with fsync of the journal, the directory and the database file before the call returns (strace: 15 database fsyncs and 38 journal fsyncs for one create, four 500-vector upserts and one delete), and the HNSW files are written every sync_threshold=1000 vectors and rebuilt from the SQLite log after a crash; measured with kill -9 and a restart: all 50, 300 and 2,000 vectors were still stored and searchable.", "Outside load when searching began: 0.17 CPUs on average over the last 1 s and 0.17 over the last few seconds (limit 0.3), a quiet box. The load average 2.97 3.49 3.29 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 23,289 searches in 30.0 s, default@1 14,060 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 11,651 searches (0 failed) at 8 searchers; QPS of the last windows 776, 778, 778 (windows of at least 2 s and 100 searches); trial of 2,339 searches in 3.0 s at 8 searchers: 778 QPS against the settled 778 QPS, 0% apart (limit 10%); the timed pass 777 QPS, 0% from the settled 778 QPS (limit 10%); default@1: warm-up 15.0 s and 7,034 searches (0 failed) at 1 searcher; p50 of the last windows 2.096, 2.104, 2.112 ms (windows of at least 2 s and 100 searches); trial of 1,411 searches in 3.0 s at 1 searcher: p50 2.101 ms against the settled p50 2.104 ms, 0% apart (limit 10%); the timed pass p50 2.105 ms, 0% from the settled p50 2.104 ms (limit 10% or 0.1 ms; 0.001 ms).", "Pass default@8 after rehearsal: 2026-10-06T14:13:38.370Z to 2026-10-06T14:13:58.378Z (20.0 s), 15,547 searches, 0 failed, p50 10.225 ms, mean 10.292 ms, p99 12.140 ms, 777.1 QPS (1000/QPS 1.287 ms).", "Pass default@1 after default@8: 2026-10-06T14:14:16.387Z to 2026-10-06T14:14:36.388Z (20.0 s), 9,379 searches, 0 failed, p50 2.105 ms, mean 2.131 ms, p99 2.653 ms, 468.9 QPS (1000/QPS 2.133 ms).", "Passes in the order run: default@8, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 59,784 sent, 0 failed (warmupErrors). Index after the searches: ready, ? of 524 indexed (Chroma exposes no index status (indexing_status endpoint: HTTP 500 {\u0022error\u0022:\u0022InternalError\u0022,\u0022message\u0022:\u0022Method scout_logs is not implemented\u0022}), so indexed vectors are not reported and ready only means the two counts agree to within 1 percent: count endpoint 524, a vector search asking for 524 results returned 524).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-chroma (94961ec20450) on network gvb-chroma_default at container-ip:8000, not through docker-proxy localhost:8000. Open after the passes: container-ip:8000 x8 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-chroma: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-chroma: 9 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3492.0/3492.2, performance, outside load 0.16 before the warm-up, 0.15 during, client CPU 0.480 ms per search; default@1 3492/3492 MHz, average 3492.0/3492.5, performance, outside load 0.16 before the warm-up, 0.16 during, client CPU 0.480 ms per search." ] }, { "name": "typesense", "engine": "Typesense 30.2 (vector search)", "index": "HNSW float32 (hnswlib), m=16, ef_construction=128, cosine; search k=top, ef=100; exact mode = filter ordinal:\u003E=0 with flat_search_cutoff; index held in memory", "hosting": "compose", "searchSettings": { "ef": "100", "ef_construction": "128", "k": "top", "m": "16" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.525169, "rowsPerSecond": 997.7740498772777, "countMatchSeconds": 0.0019899, "indexSeconds": 0.0143238, "indexNote": "nothing to build, Typesense inserts into the HNSW graph while writing; ready after 0.0 s: HNSW graph returns 524 of 524 vectors, no writes queued: num_documents 524, pending_write_batches 0, health ok, unfiltered vector query with k above the count returned 524; the graph is built in memory during each write, so there is no later build step" }, "search": { "countInTarget": 524, "latencySamples": 4705, "p50Ms": 4.1819, "p95Ms": 4.59562, "p99Ms": 5.536804000000002, "qps": { "8": 574.9689394061363, "1": 235.24331908973784 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 10477, "exactP50Ms": 5.6595, "exactP95Ms": 6.0673200000000005, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 0.5918, "1": 0.5192 } }, "ram": { "bytes": 120166809, "text": "114.6 MiB (docker stats)" }, "disk": { "bytes": 9879682, "text": "9.42 MiB added by this load (266.32 MiB (whole engine data folder), 256.9 MiB before)" }, "passOrder": [ "default@8", "default@1", "exact" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "HNSW graph returns 524 of 524 vectors, no writes queued: num_documents 524, pending_write_batches 0, health ok, unfiltered vector query with k above the count returned 524; the graph is built in memory during each write, so there is no later build step" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "HNSW graph returns 524 of 524 vectors, no writes queued: num_documents 524, pending_write_batches 0, health ok, unfiltered vector query with k above the count returned 524; the graph is built in memory during each write, so there is no later build step" } }, "durability": "Every acknowledged write is appended to Typesense\u0027s raft log and fsynced before the answer: braft raft_sync=true with raft_sync_policy=0 (sync immediately), read from the running container\u0027s brpc /flags page on 2026-10-04. A restart replays the log from the last snapshot (typesense.compose.yaml sets TYPESENSE_SNAPSHOT_INTERVAL_SECONDS=300) and rebuilds the in-memory HNSW graph, so by those settings a crash loses no acknowledged write, but the restart is slow after a big load. This is read from the settings, not shown by pulling power. One node and no replicas, so a lost disk loses the data.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started typesense.compose.yaml with every container created on CPUs 2-3,6-7: gvb-typesense cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [typesense.compose.yaml 96a165f62566]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (HNSW graph returns 524 of 524 vectors, no writes queued: num_documents 524, pending_write_batches 0, health ok, unfiltered vector query with k above the count returned 524; the graph is built in memory during each write, so there is no later build step). Durability: Every acknowledged write is appended to Typesense\u0027s raft log and fsynced before the answer: braft raft_sync=true with raft_sync_policy=0 (sync immediately), read from the running container\u0027s brpc /flags page on 2026-10-04. A restart replays the log from the last snapshot (typesense.compose.yaml sets TYPESENSE_SNAPSHOT_INTERVAL_SECONDS=300) and rebuilds the in-memory HNSW graph, so by those settings a crash loses no acknowledged write, but the restart is slow after a big load. This is read from the settings, not shown by pulling power. One node and no replicas, so a lost disk loses the data..", "Outside load when searching began: 0.28 CPUs on average over the last 5 s and 0.28 over the last few seconds (limit 0.3), a quiet box. The load average 3.02 3.17 3.19 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 17,268 searches in 30.0 s, default@1 7,062 searches in 30.0 s, exact 5,230 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 8,610 searches (0 failed) at 8 searchers; QPS of the last windows 573, 574, 574 (windows of at least 2 s and 100 searches); trial of 1,724 searches in 3.0 s at 8 searchers: 573 QPS against the settled 574 QPS, 0% apart (limit 10%); the timed pass 575 QPS, 0% from the settled 574 QPS (limit 10%); default@1: warm-up 15.0 s and 3,537 searches (0 failed) at 1 searcher; p50 of the last windows 4.176, 4.195, 4.177 ms (windows of at least 2 s and 100 searches); trial of 707 searches in 3.0 s at 1 searcher: p50 4.192 ms against the settled p50 4.177 ms, 0% apart (limit 10%); the timed pass p50 4.182 ms, 0% from the settled p50 4.177 ms (limit 10% or 0.1 ms; 0.005 ms); exact: warm-up 15.0 s and 2,613 searches (0 failed) at 1 searcher; p50 of the last windows 5.659, 5.644, 5.672 ms (windows of at least 2 s and 100 searches); trial of 525 searches in 3.0 s at 1 searcher: p50 5.653 ms against the settled p50 5.659 ms, 0% apart (limit 10%); the timed pass p50 5.660 ms, 0% from the settled p50 5.659 ms (limit 10% or 0.1 ms; 0.000 ms).", "Pass default@8 after rehearsal: 2026-10-06T14:16:37.737Z to 2026-10-06T14:16:57.743Z (20.0 s), 11,503 searches, 0 failed, p50 13.335 ms, mean 13.910 ms, p99 25.982 ms, 575.0 QPS (1000/QPS 1.739 ms).", "Pass default@1 after default@8: 2026-10-06T14:17:15.753Z to 2026-10-06T14:17:35.754Z (20.0 s), 4,705 searches, 0 failed, p50 4.182 ms, mean 4.249 ms, p99 5.537 ms, 235.2 QPS (1000/QPS 4.251 ms).", "Pass exact after default@1: 2026-10-06T14:17:53.768Z to 2026-10-06T14:18:53.772Z (60.0 s), 10,477 searches, 0 failed, p50 5.660 ms, mean 5.725 ms, p99 6.777 ms, 174.6 QPS (1000/QPS 5.727 ms).", "Passes in the order run: default@8, default@1, exact; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 47,276 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (HNSW graph returns 524 of 524 vectors, no writes queued: num_documents 524, pending_write_batches 0, health ok, unfiltered vector query with k above the count returned 524; the graph is built in memory during each write, so there is no later build step).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-typesense (77eb2dcbee94) on network engines_default at container-ip:8108, not through docker-proxy localhost:8108. Open after the passes: container-ip:8108 x1 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-typesense: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-typesense: 303 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3492.0/3492.4, performance, outside load 0.21 before the warm-up, 0.04 during, client CPU 0.592 ms per search; default@1 3492/3492 MHz, average 3492.5/3492.9, performance, outside load 0.11 before the warm-up, 0.22 during, client CPU 0.519 ms per search; exact 3492/3492 MHz, average 3492.5/3492.8, performance, outside load 0.15 before the warm-up, 0.21 during, client CPU 0.541 ms per search." ] }, { "name": "sql-diskann", "engine": "Microsoft SQL Server 2025 (RTM-CU9) (KB5122048) - 17.0.5005.3 (X64), Enterprise Developer Edition (64-bit), in container gvb-mssql (image mcr.microsoft.com/mssql/server:2025-CU9-ubuntu-24.04@sha256:2b5b581621126574f3d1f75e78d3eebe8d05aedb59ad0cfdf9aa42cb0634d726; cpuset 2-3,6-7 from its creation; SQL Server counts 4 CPU(s) and runs 4 visible scheduler(s), affinity AUTO)", "index": "DiskANN (preview) via VECTOR_SEARCH, cosine, build {\u0022StartId\u0022:\u0022306\u0022, \u0022L\u0022:\u002248\u0022, \u0022M\u0022:\u00228\u0022, \u0022R\u0022:\u002248\u0022}; the exact mode scans the same table", "hosting": "compose", "searchSettings": { "L": "48", "M": "8", "R": "48", "StartId": "306" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.8486274, "rowsPerSecond": 617.4676895890941, "countMatchSeconds": 0.0037824, "indexSeconds": 4.1916987, "indexNote": "copied 524 rows to dbo.gvb_gvbbench_eshoponweb_ann (INT key) and built DiskANN, parameters {\u0022StartId\u0022:\u0022306\u0022, \u0022L\u0022:\u002248\u0022, \u0022M\u0022:\u00228\u0022, \u0022R\u0022:\u002248\u0022}, graph covers 524 rows" }, "search": { "countInTarget": 524, "latencySamples": 5430, "p50Ms": 3.5957, "p95Ms": 4.063155, "p99Ms": 4.676517, "qps": { "1": 271.4620645337917, "8": 690.6856551446306 }, "recall": 0.9650000000000002, "ndcg": 0.5256995680516536, "errors": 0, "exactQueries": 15660, "exactP50Ms": 3.7399, "exactP95Ms": 4.251105, "exactRecall": 1, "clientCpuMsPerSearch": { "1": 1.0229, "8": 1.0946 } }, "ram": { "bytes": 556374425, "text": "530.6 MiB (docker stats)" }, "disk": { "bytes": 4734976, "text": "4.52 MiB (table and its indexes, reserved pages)" }, "passOrder": [ "default@1", "default@8", "exact" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "DiskANN index built and used. sys.vector_indexes: vix_gvb_gvbbench_eshoponweb on dbo.gvb_gvbbench_eshoponweb_ann in GvbBenchDiskAnn, DiskANN, COSINE, enabled, build {\u0022StartId\u0022:\u0022306\u0022, \u0022L\u0022:\u002248\u0022, \u0022M\u0022:\u00228\u0022, \u0022R\u0022:\u002248\u0022}; graph table rows 524 of 524 table rows. Real VECTOR_SEARCH plan: Vector Index Seek on index vix_gvb_gvbbench_eshoponweb of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann (IndexKind DiskANN), 10 rows returned, 10 hits. The exact mode searches the SAME table (dbo.gvb_gvbbench_eshoponweb_ann, database GvbBenchDiskAnn) and its real plan is: Clustered Index Scan of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann through PK_gvb_gvbbench_eshoponweb_ann (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "DiskANN index built and used. sys.vector_indexes: vix_gvb_gvbbench_eshoponweb on dbo.gvb_gvbbench_eshoponweb_ann in GvbBenchDiskAnn, DiskANN, COSINE, enabled, build {\u0022StartId\u0022:\u0022306\u0022, \u0022L\u0022:\u002248\u0022, \u0022M\u0022:\u00228\u0022, \u0022R\u0022:\u002248\u0022}; graph table rows 524 of 524 table rows. Real VECTOR_SEARCH plan: Vector Index Seek on index vix_gvb_gvbbench_eshoponweb of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann (IndexKind DiskANN), 10 rows returned, 10 hits. The exact mode searches the SAME table (dbo.gvb_gvbbench_eshoponweb_ann, database GvbBenchDiskAnn) and its real plan is: Clustered Index Scan of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann through PK_gvb_gvbbench_eshoponweb_ann (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits" } }, "durability": "A commit returns after its transaction-log records are written to disk (SQL Server write-ahead logging; delayed durability is DISABLED); a database created here copies the model database: recovery model FULL, page_verify CHECKSUM; no global trace flags are enabled; mssql.conf of container gvb-mssql (/var/opt/mssql/mssql.conf, on the host at ~/gvb-data/engines/mssql/mssql.conf) does not exist, so it sets: nothing; it has no [control] or [traceflag] entry, so SQL Server\u0027s own Linux defaults for flushing writes apply (Microsoft\u0027s Linux performance guide names trace flag 3982 as that default; read from the guide, not tested here). Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked. Container settings from its environment (names only): MSSQL_AGENT_ENABLED, MSSQL_MEMORY_LIMIT_MB, MSSQL_PID, MSSQL_RPC_PORT, MSSQL_SA_PASSWORD.", "notes": [ "Started by run-all for this measurement with every container created on CPUs 2-3,6-7 (started mssql.compose.yaml with every container created on CPUs 2-3,6-7: gvb-mssql cpuset 2-3,6-7 (read back from docker inspect)); stopped afterwards.", "Engine files (SHA-256, first 12 hex digits): [mssql.compose.yaml b7359663cc15]. This run created the engine from these files.", "Index after the load: ready, 524 of 524 indexed (DiskANN index built and used. sys.vector_indexes: vix_gvb_gvbbench_eshoponweb on dbo.gvb_gvbbench_eshoponweb_ann in GvbBenchDiskAnn, DiskANN, COSINE, enabled, build {\u0022StartId\u0022:\u0022306\u0022, \u0022L\u0022:\u002248\u0022, \u0022M\u0022:\u00228\u0022, \u0022R\u0022:\u002248\u0022}; graph table rows 524 of 524 table rows. Real VECTOR_SEARCH plan: Vector Index Seek on index vix_gvb_gvbbench_eshoponweb of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann (IndexKind DiskANN), 10 rows returned, 10 hits. The exact mode searches the SAME table (dbo.gvb_gvbbench_eshoponweb_ann, database GvbBenchDiskAnn) and its real plan is: Clustered Index Scan of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann through PK_gvb_gvbbench_eshoponweb_ann (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits). Durability: A commit returns after its transaction-log records are written to disk (SQL Server write-ahead logging; delayed durability is DISABLED); a database created here copies the model database: recovery model FULL, page_verify CHECKSUM; no global trace flags are enabled; mssql.conf of container gvb-mssql (/var/opt/mssql/mssql.conf, on the host at ~/gvb-data/engines/mssql/mssql.conf) does not exist, so it sets: nothing; it has no [control] or [traceflag] entry, so SQL Server\u0027s own Linux defaults for flushing writes apply (Microsoft\u0027s Linux performance guide names trace flag 3982 as that default; read from the guide, not tested here). Not tested by cutting power; whether the disk\u0027s own write cache reaches the media was not checked. Container settings from its environment (names only): MSSQL_AGENT_ENABLED, MSSQL_MEMORY_LIMIT_MB, MSSQL_PID, MSSQL_RPC_PORT, MSSQL_SA_PASSWORD..", "Outside load when searching began: 0.25 CPUs on average over the last 6 s and 0.24 over the last few seconds (limit 0.3), a quiet box. The load average 1.80 2.85 3.09 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@1 8,105 searches in 30.0 s, default@8 20,297 searches in 30.0 s, exact 7,826 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@1: warm-up 15.0 s and 4,091 searches (0 failed) at 1 searcher; p50 of the last windows 3.629, 3.617, 3.596 ms (windows of at least 2 s and 100 searches); trial of 824 searches in 3.0 s at 1 searcher: p50 3.605 ms against the settled p50 3.617 ms, 0% apart (limit 10%); the timed pass p50 3.596 ms, 1% from the settled p50 3.617 ms (limit 10% or 0.1 ms; 0.022 ms); default@8: warm-up 15.0 s and 9,534 searches (0 failed) at 8 searchers; QPS of the last windows 642, 637, 630 (windows of at least 2 s and 100 searches); trial of 2,044 searches in 3.0 s at 8 searchers: 679 QPS against the settled 637 QPS, 6% apart (limit 10%); the timed pass 691 QPS, 8% from the settled 637 QPS (limit 10%); exact: warm-up 15.0 s and 3,908 searches (0 failed) at 1 searcher; p50 of the last windows 3.756, 3.779, 3.784 ms (windows of at least 2 s and 100 searches); trial of 788 searches in 3.0 s at 1 searcher: p50 3.741 ms against the settled p50 3.779 ms, 1% apart (limit 10%); the timed pass p50 3.740 ms, 1% from the settled p50 3.779 ms (limit 10% or 0.1 ms; 0.039 ms).", "Pass default@1 after rehearsal: 2026-10-06T14:20:57.685Z to 2026-10-06T14:21:17.688Z (20.0 s), 5,430 searches, 0 failed, p50 3.596 ms, mean 3.681 ms, p99 4.677 ms, 271.5 QPS (1000/QPS 3.684 ms).", "Pass default@8 after default@1: 2026-10-06T14:21:35.716Z to 2026-10-06T14:21:55.725Z (20.0 s), 13,820 searches, 0 failed, p50 13.547 ms, mean 11.577 ms, p99 20.702 ms, 690.7 QPS (1000/QPS 1.448 ms).", "Pass exact after default@8: 2026-10-06T14:22:13.735Z to 2026-10-06T14:23:13.735Z (60.0 s), 15,660 searches, 0 failed, p50 3.740 ms, mean 3.828 ms, p99 4.950 ms, 261.0 QPS (1000/QPS 3.831 ms).", "Passes in the order run: default@1, default@8, exact; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 57,417 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (DiskANN index built and used. sys.vector_indexes: vix_gvb_gvbbench_eshoponweb on dbo.gvb_gvbbench_eshoponweb_ann in GvbBenchDiskAnn, DiskANN, COSINE, enabled, build {\u0022StartId\u0022:\u0022306\u0022, \u0022L\u0022:\u002248\u0022, \u0022M\u0022:\u00228\u0022, \u0022R\u0022:\u002248\u0022}; graph table rows 524 of 524 table rows. Real VECTOR_SEARCH plan: Vector Index Seek on index vix_gvb_gvbbench_eshoponweb of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann (IndexKind DiskANN), 10 rows returned, 10 hits. The exact mode searches the SAME table (dbo.gvb_gvbbench_eshoponweb_ann, database GvbBenchDiskAnn) and its real plan is: Clustered Index Scan of GvbBenchDiskAnn.gvb_gvbbench_eshoponweb_ann through PK_gvb_gvbbench_eshoponweb_ann (IndexKind Clustered), 524 rows read, no vector index operator, 10 hits).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: container address on its Docker network, not the published port (no docker-proxy): container gvb-mssql (39d172c6c285) on network gvb-mssql_default at container-ip:1433, not through docker-proxy localhost:14330. Open after the passes: container-ip:1433 x9 (this target).", "CPU pinning: cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which), CPUs 2-3,6-7. Changed: container gvb-mssql: already on CPUs 2-3,6-7 (its cpuset since creation), not changed. Read back: container gvb-mssql: 85 thread(s) on CPUs 2-3,6-7 (every thread of the container).", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@1 3492/3492 MHz, average 3492.4/3492.5, performance, outside load 0.21 before the warm-up, 0.21 during, client CPU 1.023 ms per search; default@8 3492/3492 MHz, average 3492.0/3492.3, performance, outside load 0.21 before the warm-up, 0.18 during, client CPU 1.095 ms per search; exact 3492/3492 MHz, average 3492.3/3492.4, performance, outside load 0.22 before the warm-up, 0.21 during, client CPU 1.161 ms per search." ] }, { "name": "duckdb", "engine": "DuckDB 1.5.6 \u002B vss b833341 (HNSW, embedded, in process)", "index": "HNSW (vss extension) FLOAT[n] metric=cosine m=16 ef_construction=128, ef_search=100 per connection, persistent (hnsw_enable_experimental_persistence=true, checkpoint_threshold=256MB); exact mode = array_cosine_similarity sequential scan; score = 1 - cosine distance; searches run concurrently, one connection per searcher (opened as searchers arrive, at most 32), writes run one at a time and never overlap a search", "hosting": "embedded", "searchSettings": { "checkpoint_threshold": "256MB", "ef_construction": "128", "ef_search": "100", "hnsw_enable_experimental_persistence": "true", "m": "16", "metric": "cosine" }, "load": { "rows": 524, "batch": 1000, "upsertSeconds": 0.3431596, "rowsPerSecond": 1526.9862769393599, "countMatchSeconds": 0.0029818, "indexSeconds": 0.0138298, "indexNote": "HNSW is maintained inside every transaction, nothing to build; confirmed in 0.01 s: duckdb_indexes() lists gvb_gvbbench_eshoponweb_hnsw, pragma_hnsw_index_info() counts 524 vectors of 524 rows, EXPLAIN of the default search shows HNSW_INDEX_SCAN on it = True" }, "search": { "countInTarget": 524, "latencySamples": 5393, "p50Ms": 3.6669, "p95Ms": 4.05804, "p99Ms": 5.2629399999999995, "qps": { "8": 576.0922659375863, "1": 269.63042483115726 }, "recall": 1, "ndcg": 0.5181699774768911, "errors": 0, "exactQueries": 13956, "exactP50Ms": 4.213850000000001, "exactP95Ms": 4.99935, "exactRecall": 1, "clientCpuMsPerSearch": { "8": 6.91, "1": 5.0232 } }, "ram": { "text": "in this process, not measured" }, "disk": { "bytes": 3040286, "text": "2.9 MiB added by this load (1.13 GiB (engine data folder), 1.13 GiB before)" }, "passOrder": [ "default@8", "exact", "default@1" ], "warmupErrors": 0, "indexState": { "afterLoad": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "duckdb_indexes() lists gvb_gvbbench_eshoponweb_hnsw, pragma_hnsw_index_info() counts 524 vectors of 524 rows, EXPLAIN of the default search shows HNSW_INDEX_SCAN on it = True" }, "afterSearch": { "ready": true, "indexedVectors": 524, "totalVectors": 524, "detail": "duckdb_indexes() lists gvb_gvbbench_eshoponweb_hnsw, pragma_hnsw_index_info() counts 524 vectors of 524 rows, EXPLAIN of the default search shows HNSW_INDEX_SCAN on it = True" } }, "durability": "DuckDB\u0027s write-ahead log is fsynced at every commit and no DuckDbSink setting changes that (checkpoint_threshold=256MB only spaces out the checkpoints that write the database file; measured with strace: 42 WAL fsyncs for 40 single-record commits plus 2 setup statements), so an operating-system crash or power cut loses no committed row; measured with kill -9 right after the last commit and a reopen: all 50, 300 and 2,000 rows and an HNSW index that counted the same number were recovered from the log; not tested and documented by DuckDB: the HNSW index is file-backed only through hnsw_enable_experimental_persistence = true, which this sink turns on, and WAL recovery for such custom indexes is not complete, so a crash during a checkpoint or a later commit can damage the index while the rows survive", "notes": [ "Index after the load: ready, 524 of 524 indexed (duckdb_indexes() lists gvb_gvbbench_eshoponweb_hnsw, pragma_hnsw_index_info() counts 524 vectors of 524 rows, EXPLAIN of the default search shows HNSW_INDEX_SCAN on it = True). Durability: DuckDB\u0027s write-ahead log is fsynced at every commit and no DuckDbSink setting changes that (checkpoint_threshold=256MB only spaces out the checkpoints that write the database file; measured with strace: 42 WAL fsyncs for 40 single-record commits plus 2 setup statements), so an operating-system crash or power cut loses no committed row; measured with kill -9 right after the last commit and a reopen: all 50, 300 and 2,000 rows and an HNSW index that counted the same number were recovered from the log; not tested and documented by DuckDB: the HNSW index is file-backed only through hnsw_enable_experimental_persistence = true, which this sink turns on, and WAL recovery for such custom indexes is not complete, so a crash during a checkpoint or a later commit can damage the index while the rows survive.", "Outside load when searching began: 0.12 CPUs on average over the last 1 s and 0.12 over the last few seconds (limit 0.3), a quiet box. The load average 1.80 2.58 2.95 (1/5/15 min; it also counts this benchmark\u0027s own client and engines, so it is recorded, not judged).", "Rehearsal before any timed pass, untimed, every pass type at its own concurrency for at least 30 s through the same code the timed passes use: default@8 17,526 searches in 30.0 s, exact 7,014 searches in 30.0 s, default@1 8,094 searches in 30.0 s.", "Settle: settled before every timed pass, each by its own warm-up at its own concurrency and a trial of the same pass within 10% of the warm-up\u0027s settled figure, and the timed pass itself within 10% of it. default@8: warm-up 15.0 s and 8,656 searches (0 failed) at 8 searchers; QPS of the last windows 574, 579, 576 (windows of at least 2 s and 100 searches); trial of 1,744 searches in 3.0 s at 8 searchers: 580 QPS against the settled 576 QPS, 1% apart (limit 10%); the timed pass 576 QPS, 0% from the settled 576 QPS (limit 10%); exact: warm-up 15.0 s and 3,517 searches (0 failed) at 1 searcher; p50 of the last windows 4.177, 4.166, 4.143 ms (windows of at least 2 s and 100 searches); trial of 698 searches in 3.0 s at 1 searcher: p50 4.206 ms against the settled p50 4.166 ms, 1% apart (limit 10%); the timed pass p50 4.214 ms, 1% from the settled p50 4.166 ms (limit 10% or 0.1 ms; 0.048 ms); default@1: warm-up 15.0 s and 4,048 searches (0 failed) at 1 searcher; p50 of the last windows 3.676, 3.651, 3.667 ms (windows of at least 2 s and 100 searches); trial of 806 searches in 3.0 s at 1 searcher: p50 3.670 ms against the settled p50 3.667 ms, 0% apart (limit 10%); the timed pass p50 3.667 ms, 0% from the settled p50 3.667 ms (limit 10% or 0.1 ms; 0.000 ms).", "Pass default@8 after rehearsal: 2026-10-06T14:25:04.124Z to 2026-10-06T14:25:24.130Z (20.0 s), 11,525 searches, 0 failed, p50 13.208 ms, mean 13.878 ms, p99 28.468 ms, 576.1 QPS (1000/QPS 1.736 ms).", "Pass exact after default@8: 2026-10-06T14:25:42.144Z to 2026-10-06T14:26:42.148Z (60.0 s), 13,956 searches, 0 failed, p50 4.214 ms, mean 4.296 ms, p99 5.894 ms, 232.6 QPS (1000/QPS 4.300 ms).", "Pass default@1 after exact: 2026-10-06T14:27:00.165Z to 2026-10-06T14:27:20.167Z (20.0 s), 5,393 searches, 0 failed, p50 3.667 ms, mean 3.706 ms, p99 5.263 ms, 269.6 QPS (1000/QPS 3.709 ms).", "Passes in the order run: default@8, exact, default@1; each after its own warm-up. Untimed searches (rehearsals, per-pass warm-ups, settle checks and extensions): 52,103 sent, 0 failed (warmupErrors). Index after the searches: ready, 524 of 524 indexed (duckdb_indexes() lists gvb_gvbbench_eshoponweb_hnsw, pragma_hnsw_index_info() counts 524 vectors of 524 rows, EXPLAIN of the default search shows HNSW_INDEX_SCAN on it = True).", "Benchmark copy gvbbench_eshoponweb dropped afterwards.", "Connection: in this process (embedded), no network: no network address. No open connection seen (no open TCP connection was found).", "CPU pinning: none: an embedded engine runs inside the benchmark client process, so it shares the client CPUs, CPUs 0-1,4-5.", "Clock per pass (median MHz of engine CPUs / client CPUs, their average MHz, governor, CPUs busy outside the benchmark, client CPU per search; the clock was pinned at 3500 MHz (turbo off), and a pass whose average on either group is more than 1% off it is flagged): default@8 3492/3492 MHz, average 3492.3/3492.0, performance, outside load 0.21 before the warm-up, 0.21 during, client CPU 6.910 ms per search; exact 3492/3492 MHz, average 3492.8/3492.1, performance, outside load 0.20 before the warm-up, 0.22 during, client CPU 6.269 ms per search; default@1 3492/3492 MHz, average 3492.9/3492.1, performance, outside load 0.22 before the warm-up, 0.18 during, client CPU 5.023 ms per search." ] } ], "conditions": { "buildConfiguration": "Release", "jitOptimizerDisabled": false, "dotNet": ".NET 10.0.12", "runtimeVersion": "10.0.12", "serverGc": false, "machineControl": "on", "governor": "performance", "governorBefore": "schedutil", "governorAtEnd": "performance", "governorAfterRestore": "schedutil", "logicalCpus": 8, "threadSiblings": [ "0,4", "1,5", "2,6", "3,7" ], "clientCpus": "0-1,4-5", "engineCpus": "2-3,6-7", "clientPinning": "client pid 146444: 23 thread(s) on CPUs 0-1,4-5", "warmupSearches": 20, "exactSeconds": 60, "search": { "top": 10, "queries": "golden", "concurrency": [ 1, 8 ], "secondsPerLevel": 20, "warmupSearches": 20, "exactSeconds": 60, "searchTimeoutSeconds": 120, "batch": 1000, "seed": 701 }, "busyRule": "right before each pass\u0027s warm-up the run waits, up to 10 min, while processes outside the benchmark (everything but this client and the engine under test\u0027s cgroups) use more than 0.3 CPUs on average over the last 60 s or the last 5 s, neither window reaching back past the start of the target or of its engine (so the engine\u0027s own start-up is not outside work); if the box does not clear the pass runs anyway, flagged \u0027busy box\u0027, as is a pass whose own outside load is above the limit", "engines": [ { "target": "clickhouse", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-clickhouse: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-clickhouse: 680 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of clickhouse.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "vespa", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-vespa: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-vespa: 215 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of vespa.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "oracle", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-oracle: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-oracle: 92 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of oracle.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "elasticsearch", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-elasticsearch: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-elasticsearch: 75 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of elasticsearch.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "sql", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-mssql: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-mssql: 86 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of mssql.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "pgvector", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-pgvector: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-pgvector: 6 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of pgvector.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "sqlitevec", "hosting": "embedded", "method": "none: an embedded engine runs inside the benchmark client process, so it shares the client CPUs", "cpus": "0-1,4-5", "changes": [], "verified": [], "problems": [], "restored": [] }, { "target": "qdrant", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-qdrant: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-qdrant: 24 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of qdrant.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "opensearch", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-opensearch: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-opensearch: 55 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of opensearch.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "qdrant-hnsw", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-qdrant: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-qdrant: 24 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of qdrant.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "redis", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-redis: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-redis: 16 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of redis.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "milvus", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-milvus: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-milvus: 24 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of milvus.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "mariadb", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvbbench-mariadb: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvbbench-mariadb: 13 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of mariadb-bench.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "weaviate", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-weaviate: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-weaviate: 9 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of weaviate.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "mongodb", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-mongodb: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-mongodb: 211 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of mongodb.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "chroma", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-chroma: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-chroma: 9 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of chroma.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "typesense", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-typesense: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-typesense: 303 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of typesense.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "sql-diskann", "hosting": "compose", "method": "cpuset from the container\u0027s creation when run-all started it on the engine CPUs, else docker update --cpuset-cpus (each container\u0027s change line says which)", "cpus": "2-3,6-7", "changes": [ "container gvb-mssql: already on CPUs 2-3,6-7 (its cpuset since creation), not changed" ], "verified": [ "container gvb-mssql: 85 thread(s) on CPUs 2-3,6-7 (every thread of the container)" ], "problems": [], "restored": [ "no container of mssql.compose.yaml is left on CPUs 2-3,6-7; nothing to put back" ] }, { "target": "duckdb", "hosting": "embedded", "method": "none: an embedded engine runs inside the benchmark client process, so it shares the client CPUs", "cpus": "0-1,4-5", "changes": [], "verified": [], "problems": [], "restored": [] } ], "passes": [ { "target": "clickhouse", "pass": "default@1", "startUtc": "2026-10-06T13:08:22.782Z", "endUtc": "2026-10-06T13:08:42.820Z", "seconds": 20.038, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3486, "median": 3492, "max": 3494 }, { "cpu": 1, "min": 3488, "median": 3492, "max": 3499 }, { "cpu": 2, "min": 3490, "median": 3492, "max": 3494 }, { "cpu": 3, "min": 3486, "median": 3492, "max": 3493 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3498 }, { "cpu": 6, "min": 3488, "median": 3492, "max": 3496 }, { "cpu": 7, "min": 3491, "median": 3492, "max": 3494 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:08:04.752Z", "quietCheckLeadSeconds": 18.03, "outsideWindowFromUtc": "2026-10-06T13:07:04.752Z", "outsideLoadAtStart": 0.244, "outsideLoadRecentAtStart": 0.221, "outsideLoadDuring": 0.251, "accountingMismatch": false, "searches": 3832, "clientCpuSeconds": 3.3453, "clientCpuMsPerSearch": 0.873, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:08:22.784Z to 2026-10-06T13:08:42.787Z, 80 samples, divided by the pass\u0027s 3832 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "clickhouse", "pass": "default@8", "startUtc": "2026-10-06T13:09:53.945Z", "endUtc": "2026-10-06T13:10:13.960Z", "seconds": 20.015, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3501 }, { "cpu": 1, "min": 3489, "median": 3492, "max": 3497 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3485, "median": 3492, "max": 3499 }, { "cpu": 5, "min": 3489, "median": 3492, "max": 3498 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:09:38.931Z", "quietCheckLeadSeconds": 15.014, "outsideWindowFromUtc": "2026-10-06T13:08:38.931Z", "outsideLoadAtStart": 0.248, "outsideLoadRecentAtStart": 0.256, "outsideLoadDuring": 0.249, "accountingMismatch": false, "searches": 7500, "clientCpuSeconds": 7.9255, "clientCpuMsPerSearch": 1.0567, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:09:53.945Z to 2026-10-06T13:10:13.956Z, 80 samples, divided by the pass\u0027s 7500 completed searches", "waitedSeconds": 3, "busyBox": false }, { "target": "clickhouse", "pass": "exact", "startUtc": "2026-10-06T13:10:31.966Z", "endUtc": "2026-10-06T13:11:31.976Z", "seconds": 60.01, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3488, "median": 3492, "max": 3497 }, { "cpu": 1, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3450, "median": 3492, "max": 3501 }, { "cpu": 3, "min": 3480, "median": 3492, "max": 3494 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3478, "median": 3492, "max": 3504 }, { "cpu": 6, "min": 3444, "median": 3492, "max": 3503 }, { "cpu": 7, "min": 3488, "median": 3492, "max": 3495 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:10:13.961Z", "quietCheckLeadSeconds": 18.006, "outsideWindowFromUtc": "2026-10-06T13:09:13.961Z", "outsideLoadAtStart": 0.253, "outsideLoadRecentAtStart": 0.254, "outsideLoadDuring": 0.249, "accountingMismatch": false, "searches": 8686, "clientCpuSeconds": 7.8269, "clientCpuMsPerSearch": 0.9011, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:10:31.966Z to 2026-10-06T13:11:31.968Z, 240 samples, divided by the pass\u0027s 8686 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "vespa", "pass": "exact", "startUtc": "2026-10-06T13:14:24.184Z", "endUtc": "2026-10-06T13:15:24.216Z", "seconds": 60.032, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3487, "median": 3492, "max": 3526 }, { "cpu": 1, "min": 3489, "median": 3492, "max": 3502 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3502 }, { "cpu": 5, "min": 3481, "median": 3492, "max": 3501 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3491, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:14:06.179Z", "quietCheckLeadSeconds": 18.005, "outsideWindowFromUtc": "2026-10-06T13:13:06.179Z", "outsideLoadAtStart": 0.224, "outsideLoadRecentAtStart": 0.267, "outsideLoadDuring": 0.19, "accountingMismatch": false, "searches": 29472, "clientCpuSeconds": 27.4525, "clientCpuMsPerSearch": 0.9315, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:14:24.184Z to 2026-10-06T13:15:24.184Z, 240 samples, divided by the pass\u0027s 29472 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "vespa", "pass": "default@1", "startUtc": "2026-10-06T13:15:42.220Z", "endUtc": "2026-10-06T13:16:02.230Z", "seconds": 20.01, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 1, "min": 3467, "median": 3492, "max": 3498 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3497 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3499 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:15:24.217Z", "quietCheckLeadSeconds": 18.002, "outsideWindowFromUtc": "2026-10-06T13:14:24.217Z", "outsideLoadAtStart": 0.19, "outsideLoadRecentAtStart": 0.176, "outsideLoadDuring": 0.189, "accountingMismatch": false, "searches": 9831, "clientCpuSeconds": 9.4171, "clientCpuMsPerSearch": 0.9579, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:15:42.220Z to 2026-10-06T13:16:02.220Z, 80 samples, divided by the pass\u0027s 9831 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "vespa", "pass": "default@8", "startUtc": "2026-10-06T13:17:08.249Z", "endUtc": "2026-10-06T13:17:28.263Z", "seconds": 20.014, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3496 }, { "cpu": 1, "min": 3492, "median": 3492, "max": 3497 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3504 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3498 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:16:53.245Z", "quietCheckLeadSeconds": 15.004, "outsideWindowFromUtc": "2026-10-06T13:15:53.245Z", "outsideLoadAtStart": 0.269, "outsideLoadRecentAtStart": 0.285, "outsideLoadDuring": 0.276, "accountingMismatch": false, "searches": 30530, "clientCpuSeconds": 26.8683, "clientCpuMsPerSearch": 0.8801, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:17:08.249Z to 2026-10-06T13:17:28.253Z, 80 samples, divided by the pass\u0027s 30530 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "oracle", "pass": "default@8", "startUtc": "2026-10-06T13:20:32.534Z", "endUtc": "2026-10-06T13:20:52.677Z", "seconds": 20.144, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 1196, "median": 3492, "max": 3500 }, { "cpu": 1, "min": 1198, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 1196, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 1197, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 1198, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 1197, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 1197, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 1194, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:20:17.532Z", "quietCheckLeadSeconds": 15.001, "outsideWindowFromUtc": "2026-10-06T13:19:17.532Z", "outsideLoadAtStart": 0.257, "outsideLoadRecentAtStart": 0.239, "outsideLoadDuring": 0.217, "accountingMismatch": false, "searches": 56389, "clientCpuSeconds": 36.2786, "clientCpuMsPerSearch": 0.6434, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:20:32.534Z to 2026-10-06T13:20:52.657Z, 80 samples, divided by the pass\u0027s 56389 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "oracle", "pass": "exact", "startUtc": "2026-10-06T13:21:10.680Z", "endUtc": "2026-10-06T13:22:10.704Z", "seconds": 60.023, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3408, "median": 3492, "max": 3499 }, { "cpu": 1, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3489, "median": 3492, "max": 3496 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3487, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:20:52.678Z", "quietCheckLeadSeconds": 18.002, "outsideWindowFromUtc": "2026-10-06T13:19:52.678Z", "outsideLoadAtStart": 0.257, "outsideLoadRecentAtStart": 0.178, "outsideLoadDuring": 0.227, "accountingMismatch": false, "searches": 23504, "clientCpuSeconds": 21.8395, "clientCpuMsPerSearch": 0.9292, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:21:10.680Z to 2026-10-06T13:22:10.681Z, 240 samples, divided by the pass\u0027s 23504 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "oracle", "pass": "default@1", "startUtc": "2026-10-06T13:22:28.708Z", "endUtc": "2026-10-06T13:22:48.733Z", "seconds": 20.025, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3455, "median": 3492, "max": 3494 }, { "cpu": 1, "min": 3488, "median": 3492, "max": 3493 }, { "cpu": 2, "min": 3488, "median": 3492, "max": 3495 }, { "cpu": 3, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3499 }, { "cpu": 5, "min": 3492, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3488, "median": 3492, "max": 3496 }, { "cpu": 7, "min": 3489, "median": 3492, "max": 3496 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:22:10.704Z", "quietCheckLeadSeconds": 18.004, "outsideWindowFromUtc": "2026-10-06T13:21:10.704Z", "outsideLoadAtStart": 0.227, "outsideLoadRecentAtStart": 0.259, "outsideLoadDuring": 0.232, "accountingMismatch": false, "searches": 21018, "clientCpuSeconds": 18.5005, "clientCpuMsPerSearch": 0.8802, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:22:28.708Z to 2026-10-06T13:22:48.708Z, 80 samples, divided by the pass\u0027s 21018 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "elasticsearch", "pass": "exact", "startUtc": "2026-10-06T13:25:15.144Z", "endUtc": "2026-10-06T13:26:15.193Z", "seconds": 60.049, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 2823, "median": 3492, "max": 3496 }, { "cpu": 1, "min": 3457, "median": 3492, "max": 3501 }, { "cpu": 2, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3487, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3555 }, { "cpu": 6, "min": 3438, "median": 3492, "max": 3499 }, { "cpu": 7, "min": 3393, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:24:57.140Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:23:57.140Z", "outsideLoadAtStart": 0.219, "outsideLoadRecentAtStart": 0.209, "outsideLoadDuring": 0.215, "accountingMismatch": false, "searches": 46096, "clientCpuSeconds": 21.2528, "clientCpuMsPerSearch": 0.4611, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:25:15.144Z to 2026-10-06T13:26:15.144Z, 240 samples, divided by the pass\u0027s 46096 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "elasticsearch", "pass": "default@8", "startUtc": "2026-10-06T13:26:33.198Z", "endUtc": "2026-10-06T13:26:53.227Z", "seconds": 20.029, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3501 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3498 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3496 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:26:15.194Z", "quietCheckLeadSeconds": 18.004, "outsideWindowFromUtc": "2026-10-06T13:25:15.194Z", "outsideLoadAtStart": 0.215, "outsideLoadRecentAtStart": 0.217, "outsideLoadDuring": 0.236, "accountingMismatch": false, "searches": 53431, "clientCpuSeconds": 26.5454, "clientCpuMsPerSearch": 0.4968, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:26:33.198Z to 2026-10-06T13:26:53.200Z, 80 samples, divided by the pass\u0027s 53431 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "elasticsearch", "pass": "default@1", "startUtc": "2026-10-06T13:27:11.230Z", "endUtc": "2026-10-06T13:27:31.248Z", "seconds": 20.018, "endedBy": "result line", "governor": "performance", "frequencySamples": 81, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3494 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3475, "median": 3492, "max": 3494 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3472, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3491, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:26:53.228Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:25:53.228Z", "outsideLoadAtStart": 0.241, "outsideLoadRecentAtStart": 0.234, "outsideLoadDuring": 0.21, "accountingMismatch": false, "searches": 14790, "clientCpuSeconds": 6.8504, "clientCpuMsPerSearch": 0.4632, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:27:11.230Z to 2026-10-06T13:27:31.232Z, 80 samples, divided by the pass\u0027s 14790 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sql", "pass": "default@8", "startUtc": "2026-10-06T13:29:04.896Z", "endUtc": "2026-10-06T13:29:24.911Z", "seconds": 20.014, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3440, "median": 3492, "max": 3496 }, { "cpu": 1, "min": 3488, "median": 3492, "max": 3495 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3497 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:28:46.877Z", "quietCheckLeadSeconds": 18.02, "outsideWindowFromUtc": "2026-10-06T13:27:46.877Z", "outsideLoadAtStart": 0.215, "outsideLoadRecentAtStart": 0.192, "outsideLoadDuring": 0.257, "accountingMismatch": false, "searches": 12239, "clientCpuSeconds": 14.4889, "clientCpuMsPerSearch": 1.1838, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:29:04.896Z to 2026-10-06T13:29:24.907Z, 80 samples, divided by the pass\u0027s 12239 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sql", "pass": "default@1", "startUtc": "2026-10-06T13:29:42.916Z", "endUtc": "2026-10-06T13:30:02.922Z", "seconds": 20.006, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3486, "median": 3492, "max": 3496 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3495 }, { "cpu": 2, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3502 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3487, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:29:24.911Z", "quietCheckLeadSeconds": 18.005, "outsideWindowFromUtc": "2026-10-06T13:28:24.911Z", "outsideLoadAtStart": 0.233, "outsideLoadRecentAtStart": 0.236, "outsideLoadDuring": 0.209, "accountingMismatch": false, "searches": 4820, "clientCpuSeconds": 5.8772, "clientCpuMsPerSearch": 1.2193, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:29:42.916Z to 2026-10-06T13:30:02.917Z, 80 samples, divided by the pass\u0027s 4820 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "pgvector", "pass": "exact", "startUtc": "2026-10-06T13:32:00.427Z", "endUtc": "2026-10-06T13:33:00.452Z", "seconds": 60.025, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3455, "median": 3492, "max": 3499 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3485, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3477, "median": 3492, "max": 3501 }, { "cpu": 5, "min": 3485, "median": 3492, "max": 3501 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3504 }, { "cpu": 7, "min": 3482, "median": 3492, "max": 3502 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:31:42.422Z", "quietCheckLeadSeconds": 18.005, "outsideWindowFromUtc": "2026-10-06T13:30:42.422Z", "outsideLoadAtStart": 0.218, "outsideLoadRecentAtStart": 0.218, "outsideLoadDuring": 0.196, "accountingMismatch": false, "searches": 26025, "clientCpuSeconds": 18.0154, "clientCpuMsPerSearch": 0.6922, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:32:00.427Z to 2026-10-06T13:33:00.427Z, 240 samples, divided by the pass\u0027s 26025 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "pgvector", "pass": "default@8", "startUtc": "2026-10-06T13:33:18.457Z", "endUtc": "2026-10-06T13:33:38.488Z", "seconds": 20.031, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 1, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3492 }, { "cpu": 5, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:33:00.453Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:32:00.453Z", "outsideLoadAtStart": 0.196, "outsideLoadRecentAtStart": 0.195, "outsideLoadDuring": 0.27, "accountingMismatch": false, "searches": 78682, "clientCpuSeconds": 42.2986, "clientCpuMsPerSearch": 0.5376, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:33:18.457Z to 2026-10-06T13:33:38.458Z, 80 samples, divided by the pass\u0027s 78682 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "pgvector", "pass": "default@1", "startUtc": "2026-10-06T13:33:56.492Z", "endUtc": "2026-10-06T13:34:16.526Z", "seconds": 20.035, "endedBy": "result line", "governor": "performance", "frequencySamples": 81, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3430, "median": 3492, "max": 3494 }, { "cpu": 1, "min": 3489, "median": 3492, "max": 3495 }, { "cpu": 2, "min": 3485, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3488, "median": 3492, "max": 3501 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3489, "median": 3492, "max": 3497 }, { "cpu": 6, "min": 3487, "median": 3492, "max": 3509 }, { "cpu": 7, "min": 3490, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:33:38.488Z", "quietCheckLeadSeconds": 18.004, "outsideWindowFromUtc": "2026-10-06T13:32:38.488Z", "outsideLoadAtStart": 0.244, "outsideLoadRecentAtStart": 0.246, "outsideLoadDuring": 0.183, "accountingMismatch": false, "searches": 22143, "clientCpuSeconds": 15.2296, "clientCpuMsPerSearch": 0.6878, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:33:56.492Z to 2026-10-06T13:34:16.492Z, 80 samples, divided by the pass\u0027s 22143 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sqlitevec", "pass": "default@1", "startUtc": "2026-10-06T13:36:06.299Z", "endUtc": "2026-10-06T13:36:26.310Z", "seconds": 20.01, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3486, "median": 3492, "max": 3500 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3471, "median": 3492, "max": 3537 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3501 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3548 }, { "cpu": 6, "min": 3487, "median": 3493, "max": 3526 }, { "cpu": 7, "min": 3490, "median": 3492, "max": 3502 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:35:48.298Z", "quietCheckLeadSeconds": 18.001, "outsideWindowFromUtc": "2026-10-06T13:34:48.298Z", "outsideLoadAtStart": 0.226, "outsideLoadRecentAtStart": 0.217, "outsideLoadDuring": 0.242, "accountingMismatch": false, "searches": 10198, "clientCpuSeconds": 21.0814, "clientCpuMsPerSearch": 2.0672, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:36:06.299Z to 2026-10-06T13:36:26.300Z, 80 samples, divided by the pass\u0027s 10198 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sqlitevec", "pass": "default@8", "startUtc": "2026-10-06T13:36:44.318Z", "endUtc": "2026-10-06T13:37:04.328Z", "seconds": 20.01, "endedBy": "result line", "governor": "performance", "frequencySamples": 77, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 1, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 2, "min": 3489, "median": 3492, "max": 3496 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 5, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 6, "min": 3471, "median": 3492, "max": 3498 }, { "cpu": 7, "min": 3488, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:36:26.311Z", "quietCheckLeadSeconds": 18.007, "outsideWindowFromUtc": "2026-10-06T13:35:26.311Z", "outsideLoadAtStart": 0.223, "outsideLoadRecentAtStart": 0.204, "outsideLoadDuring": 0.227, "accountingMismatch": false, "searches": 22843, "clientCpuSeconds": 79.7785, "clientCpuMsPerSearch": 3.4925, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:36:44.318Z to 2026-10-06T13:37:04.322Z, 77 samples, divided by the pass\u0027s 22843 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sqlitevec", "pass": "exact", "startUtc": "2026-10-06T13:37:22.332Z", "endUtc": "2026-10-06T13:38:22.364Z", "seconds": 60.032, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3487, "median": 3492, "max": 3500 }, { "cpu": 1, "min": 3487, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3434, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3485, "median": 3492, "max": 3503 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3587 }, { "cpu": 5, "min": 3423, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3489, "median": 3492, "max": 3504 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:37:04.328Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:36:04.328Z", "outsideLoadAtStart": 0.234, "outsideLoadRecentAtStart": 0.238, "outsideLoadDuring": 0.221, "accountingMismatch": false, "searches": 30824, "clientCpuSeconds": 62.8366, "clientCpuMsPerSearch": 2.0386, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:37:22.332Z to 2026-10-06T13:38:22.333Z, 240 samples, divided by the pass\u0027s 30824 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "qdrant", "pass": "default@1", "startUtc": "2026-10-06T13:39:48.157Z", "endUtc": "2026-10-06T13:40:08.182Z", "seconds": 20.025, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3496 }, { "cpu": 4, "min": 3427, "median": 3492, "max": 3495 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3497 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3493 }, { "cpu": 7, "min": 3491, "median": 3492, "max": 3493 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:39:30.153Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:38:30.153Z", "outsideLoadAtStart": 0.193, "outsideLoadRecentAtStart": 0.2, "outsideLoadDuring": 0.176, "accountingMismatch": false, "searches": 22224, "clientCpuSeconds": 16.3847, "clientCpuMsPerSearch": 0.7373, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:39:48.157Z to 2026-10-06T13:40:08.157Z, 80 samples, divided by the pass\u0027s 22224 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "qdrant", "pass": "default@8", "startUtc": "2026-10-06T13:40:26.187Z", "endUtc": "2026-10-06T13:40:46.214Z", "seconds": 20.027, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 1, "min": 3485, "median": 3492, "max": 3493 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3499 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:40:08.182Z", "quietCheckLeadSeconds": 18.005, "outsideWindowFromUtc": "2026-10-06T13:39:08.182Z", "outsideLoadAtStart": 0.188, "outsideLoadRecentAtStart": 0.167, "outsideLoadDuring": 0.216, "accountingMismatch": false, "searches": 65483, "clientCpuSeconds": 34.5228, "clientCpuMsPerSearch": 0.5272, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:40:26.187Z to 2026-10-06T13:40:46.189Z, 80 samples, divided by the pass\u0027s 65483 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "opensearch", "pass": "exact", "startUtc": "2026-10-06T13:43:05.836Z", "endUtc": "2026-10-06T13:44:05.858Z", "seconds": 60.022, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3469, "median": 3492, "max": 3500 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3503 }, { "cpu": 2, "min": 3479, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3425, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3444, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3488, "median": 3492, "max": 3498 }, { "cpu": 7, "min": 3489, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:42:47.831Z", "quietCheckLeadSeconds": 18.005, "outsideWindowFromUtc": "2026-10-06T13:41:47.831Z", "outsideLoadAtStart": 0.198, "outsideLoadRecentAtStart": 0.175, "outsideLoadDuring": 0.211, "accountingMismatch": false, "searches": 21439, "clientCpuSeconds": 10.1549, "clientCpuMsPerSearch": 0.4737, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:43:05.836Z to 2026-10-06T13:44:05.838Z, 240 samples, divided by the pass\u0027s 21439 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "opensearch", "pass": "default@8", "startUtc": "2026-10-06T13:44:23.868Z", "endUtc": "2026-10-06T13:44:43.880Z", "seconds": 20.012, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3498 }, { "cpu": 1, "min": 3488, "median": 3492, "max": 3496 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3497 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3498 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:44:05.859Z", "quietCheckLeadSeconds": 18.01, "outsideWindowFromUtc": "2026-10-06T13:43:05.859Z", "outsideLoadAtStart": 0.211, "outsideLoadRecentAtStart": 0.214, "outsideLoadDuring": 0.203, "accountingMismatch": false, "searches": 26932, "clientCpuSeconds": 13.8622, "clientCpuMsPerSearch": 0.5147, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:44:23.868Z to 2026-10-06T13:44:43.873Z, 80 samples, divided by the pass\u0027s 26932 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "opensearch", "pass": "default@1", "startUtc": "2026-10-06T13:45:01.885Z", "endUtc": "2026-10-06T13:45:21.896Z", "seconds": 20.011, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3493 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3480, "median": 3492, "max": 3494 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3493 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3492, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3498 }, { "cpu": 7, "min": 3485, "median": 3492, "max": 3496 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:44:43.881Z", "quietCheckLeadSeconds": 18.004, "outsideWindowFromUtc": "2026-10-06T13:43:43.881Z", "outsideLoadAtStart": 0.205, "outsideLoadRecentAtStart": 0.195, "outsideLoadDuring": 0.21, "accountingMismatch": false, "searches": 9018, "clientCpuSeconds": 4.2657, "clientCpuMsPerSearch": 0.473, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:45:01.885Z to 2026-10-06T13:45:21.887Z, 80 samples, divided by the pass\u0027s 9018 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "qdrant-hnsw", "pass": "default@1", "startUtc": "2026-10-06T13:47:20.324Z", "endUtc": "2026-10-06T13:47:40.349Z", "seconds": 20.025, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3494 }, { "cpu": 1, "min": 3492, "median": 3492, "max": 3497 }, { "cpu": 2, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3494 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3495 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3495 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3493 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:47:02.321Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:46:02.321Z", "outsideLoadAtStart": 0.199, "outsideLoadRecentAtStart": 0.145, "outsideLoadDuring": 0.166, "accountingMismatch": false, "searches": 21430, "clientCpuSeconds": 15.5808, "clientCpuMsPerSearch": 0.7271, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:47:20.324Z to 2026-10-06T13:47:40.325Z, 80 samples, divided by the pass\u0027s 21430 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "qdrant-hnsw", "pass": "default@8", "startUtc": "2026-10-06T13:47:58.352Z", "endUtc": "2026-10-06T13:48:18.378Z", "seconds": 20.025, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 1, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 5, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:47:40.349Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:46:40.349Z", "outsideLoadAtStart": 0.162, "outsideLoadRecentAtStart": 0.128, "outsideLoadDuring": 0.225, "accountingMismatch": false, "searches": 69044, "clientCpuSeconds": 35.2274, "clientCpuMsPerSearch": 0.5102, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:47:58.352Z to 2026-10-06T13:48:18.354Z, 80 samples, divided by the pass\u0027s 69044 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "qdrant-hnsw", "pass": "exact", "startUtc": "2026-10-06T13:48:36.383Z", "endUtc": "2026-10-06T13:49:36.453Z", "seconds": 60.07, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3436, "median": 3492, "max": 3496 }, { "cpu": 1, "min": 3489, "median": 3492, "max": 3529 }, { "cpu": 2, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 4, "min": 3488, "median": 3492, "max": 3495 }, { "cpu": 5, "min": 3423, "median": 3492, "max": 3496 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3494 }, { "cpu": 7, "min": 3491, "median": 3492, "max": 3493 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:48:18.378Z", "quietCheckLeadSeconds": 18.004, "outsideWindowFromUtc": "2026-10-06T13:47:18.378Z", "outsideLoadAtStart": 0.221, "outsideLoadRecentAtStart": 0.198, "outsideLoadDuring": 0.171, "accountingMismatch": false, "searches": 66103, "clientCpuSeconds": 48.2791, "clientCpuMsPerSearch": 0.7304, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:48:36.383Z to 2026-10-06T13:49:36.383Z, 240 samples, divided by the pass\u0027s 66103 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "redis", "pass": "default@8", "startUtc": "2026-10-06T13:51:46.664Z", "endUtc": "2026-10-06T13:52:06.702Z", "seconds": 20.038, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3487, "median": 3492, "max": 3493 }, { "cpu": 1, "min": 3487, "median": 3492, "max": 3497 }, { "cpu": 2, "min": 3490, "median": 3492, "max": 3496 }, { "cpu": 3, "min": 3489, "median": 3492, "max": 3496 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3493 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 6, "min": 3489, "median": 3492, "max": 3495 }, { "cpu": 7, "min": 3488, "median": 3492, "max": 3494 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:51:28.660Z", "quietCheckLeadSeconds": 18.004, "outsideWindowFromUtc": "2026-10-06T13:50:28.660Z", "outsideLoadAtStart": 0.191, "outsideLoadRecentAtStart": 0.256, "outsideLoadDuring": 0.196, "accountingMismatch": false, "searches": 102408, "clientCpuSeconds": 37.2788, "clientCpuMsPerSearch": 0.364, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:51:46.664Z to 2026-10-06T13:52:06.665Z, 80 samples, divided by the pass\u0027s 102408 completed searches", "waitedSeconds": 3, "busyBox": false }, { "target": "redis", "pass": "default@1", "startUtc": "2026-10-06T13:52:24.709Z", "endUtc": "2026-10-06T13:52:44.768Z", "seconds": 20.059, "endedBy": "result line", "governor": "performance", "frequencySamples": 81, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3488, "median": 3492, "max": 3493 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3496 }, { "cpu": 2, "min": 3489, "median": 3492, "max": 3494 }, { "cpu": 3, "min": 3490, "median": 3492, "max": 3496 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3496 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3489, "median": 3492, "max": 3501 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:52:06.703Z", "quietCheckLeadSeconds": 18.006, "outsideWindowFromUtc": "2026-10-06T13:51:06.703Z", "outsideLoadAtStart": 0.197, "outsideLoadRecentAtStart": 0.23, "outsideLoadDuring": 0.196, "accountingMismatch": false, "searches": 50539, "clientCpuSeconds": 28.7421, "clientCpuMsPerSearch": 0.5687, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:52:24.709Z to 2026-10-06T13:52:44.709Z, 80 samples, divided by the pass\u0027s 50539 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "redis", "pass": "exact", "startUtc": "2026-10-06T13:53:02.775Z", "endUtc": "2026-10-06T13:54:02.960Z", "seconds": 60.185, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3495 }, { "cpu": 2, "min": 3489, "median": 3492, "max": 3501 }, { "cpu": 3, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3486, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3482, "median": 3492, "max": 3505 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:52:44.769Z", "quietCheckLeadSeconds": 18.006, "outsideWindowFromUtc": "2026-10-06T13:51:44.769Z", "outsideLoadAtStart": 0.209, "outsideLoadRecentAtStart": 0.215, "outsideLoadDuring": 0.199, "accountingMismatch": false, "searches": 159894, "clientCpuSeconds": 90.2533, "clientCpuMsPerSearch": 0.5645, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:53:02.775Z to 2026-10-06T13:54:02.775Z, 240 samples, divided by the pass\u0027s 159894 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "milvus", "pass": "default@8", "startUtc": "2026-10-06T13:57:51.754Z", "endUtc": "2026-10-06T13:58:11.765Z", "seconds": 20.011, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3489, "median": 3492, "max": 3499 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3496 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3488, "median": 3492, "max": 3498 }, { "cpu": 5, "min": 3489, "median": 3492, "max": 3499 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3494 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3494 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:57:33.748Z", "quietCheckLeadSeconds": 18.007, "outsideWindowFromUtc": "2026-10-06T13:56:33.748Z", "outsideLoadAtStart": 0.225, "outsideLoadRecentAtStart": 0.165, "outsideLoadDuring": 0.254, "accountingMismatch": false, "searches": 25099, "clientCpuSeconds": 16.023, "clientCpuMsPerSearch": 0.6384, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:57:51.754Z to 2026-10-06T13:58:11.758Z, 80 samples, divided by the pass\u0027s 25099 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "milvus", "pass": "default@1", "startUtc": "2026-10-06T13:58:29.776Z", "endUtc": "2026-10-06T13:58:49.785Z", "seconds": 20.009, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3494 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3496 }, { "cpu": 2, "min": 3491, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3486, "median": 3492, "max": 3497 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3493 }, { "cpu": 7, "min": 3486, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T13:58:11.766Z", "quietCheckLeadSeconds": 18.011, "outsideWindowFromUtc": "2026-10-06T13:57:11.766Z", "outsideLoadAtStart": 0.234, "outsideLoadRecentAtStart": 0.231, "outsideLoadDuring": 0.194, "accountingMismatch": false, "searches": 8108, "clientCpuSeconds": 4.6103, "clientCpuMsPerSearch": 0.5686, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T13:58:29.776Z to 2026-10-06T13:58:49.777Z, 80 samples, divided by the pass\u0027s 8108 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "mariadb", "pass": "default@8", "startUtc": "2026-10-06T14:00:53.785Z", "endUtc": "2026-10-06T14:01:13.830Z", "seconds": 20.045, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3489, "median": 3492, "max": 3492 }, { "cpu": 1, "min": 3492, "median": 3492, "max": 3499 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 5, "min": 3487, "median": 3492, "max": 3502 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:00:35.782Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T13:59:35.782Z", "outsideLoadAtStart": 0.2, "outsideLoadRecentAtStart": 0.185, "outsideLoadDuring": 0.159, "accountingMismatch": false, "searches": 118236, "clientCpuSeconds": 44.4156, "clientCpuMsPerSearch": 0.3757, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:00:53.785Z to 2026-10-06T14:01:13.786Z, 80 samples, divided by the pass\u0027s 118236 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "mariadb", "pass": "default@1", "startUtc": "2026-10-06T14:01:31.833Z", "endUtc": "2026-10-06T14:01:51.867Z", "seconds": 20.034, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3493 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3490, "median": 3492, "max": 3502 }, { "cpu": 3, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3483, "median": 3492, "max": 3501 }, { "cpu": 7, "min": 3490, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:01:13.830Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T14:00:13.830Z", "outsideLoadAtStart": 0.17, "outsideLoadRecentAtStart": 0.175, "outsideLoadDuring": 0.188, "accountingMismatch": false, "searches": 30928, "clientCpuSeconds": 15.0242, "clientCpuMsPerSearch": 0.4858, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:01:31.833Z to 2026-10-06T14:01:51.834Z, 80 samples, divided by the pass\u0027s 30928 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "mariadb", "pass": "exact", "startUtc": "2026-10-06T14:02:09.870Z", "endUtc": "2026-10-06T14:03:09.906Z", "seconds": 60.036, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3488, "median": 3492, "max": 3502 }, { "cpu": 1, "min": 3486, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3486, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3285, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3475, "median": 3492, "max": 3502 }, { "cpu": 6, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3489, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:01:51.868Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T14:00:51.868Z", "outsideLoadAtStart": 0.187, "outsideLoadRecentAtStart": 0.163, "outsideLoadDuring": 0.187, "accountingMismatch": false, "searches": 35976, "clientCpuSeconds": 17.9336, "clientCpuMsPerSearch": 0.4985, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:02:09.870Z to 2026-10-06T14:03:09.872Z, 240 samples, divided by the pass\u0027s 35976 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "weaviate", "pass": "default@1", "startUtc": "2026-10-06T14:04:51.911Z", "endUtc": "2026-10-06T14:05:11.914Z", "seconds": 20.004, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3486, "median": 3492, "max": 3499 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3497 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3501 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3502 }, { "cpu": 6, "min": 3486, "median": 3492, "max": 3503 }, { "cpu": 7, "min": 3488, "median": 3492, "max": 3495 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:04:33.904Z", "quietCheckLeadSeconds": 18.007, "outsideWindowFromUtc": "2026-10-06T14:03:33.904Z", "outsideLoadAtStart": 0.204, "outsideLoadRecentAtStart": 0.172, "outsideLoadDuring": 0.204, "accountingMismatch": false, "searches": 3313, "clientCpuSeconds": 1.8957, "clientCpuMsPerSearch": 0.5722, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:04:51.911Z to 2026-10-06T14:05:11.911Z, 80 samples, divided by the pass\u0027s 3313 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "weaviate", "pass": "default@8", "startUtc": "2026-10-06T14:05:29.950Z", "endUtc": "2026-10-06T14:05:49.970Z", "seconds": 20.02, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3493 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3498 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:05:11.915Z", "quietCheckLeadSeconds": 18.034, "outsideWindowFromUtc": "2026-10-06T14:04:11.915Z", "outsideLoadAtStart": 0.205, "outsideLoadRecentAtStart": 0.23, "outsideLoadDuring": 0.227, "accountingMismatch": false, "searches": 6104, "clientCpuSeconds": 3.7677, "clientCpuMsPerSearch": 0.6173, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:05:29.950Z to 2026-10-06T14:05:49.968Z, 80 samples, divided by the pass\u0027s 6104 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "mongodb", "pass": "default@1", "startUtc": "2026-10-06T14:08:01.467Z", "endUtc": "2026-10-06T14:08:21.481Z", "seconds": 20.014, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3488, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3458, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3491, "median": 3492, "max": 3497 }, { "cpu": 6, "min": 3474, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3491, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:07:43.465Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T14:06:43.465Z", "outsideLoadAtStart": 0.264, "outsideLoadRecentAtStart": 0.257, "outsideLoadDuring": 0.293, "accountingMismatch": false, "searches": 13821, "clientCpuSeconds": 4.3101, "clientCpuMsPerSearch": 0.3118, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:08:01.467Z to 2026-10-06T14:08:21.468Z, 80 samples, divided by the pass\u0027s 13821 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "mongodb", "pass": "exact", "startUtc": "2026-10-06T14:08:40.485Z", "endUtc": "2026-10-06T14:09:40.530Z", "seconds": 60.046, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3488, "median": 3492, "max": 3496 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 2546, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3490, "median": 3492, "max": 3504 }, { "cpu": 7, "min": 3463, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:08:22.482Z", "quietCheckLeadSeconds": 18.002, "outsideWindowFromUtc": "2026-10-06T14:07:22.482Z", "outsideLoadAtStart": 0.272, "outsideLoadRecentAtStart": 0.259, "outsideLoadDuring": 0.307, "accountingMismatch": false, "searches": 47151, "clientCpuSeconds": 14.322, "clientCpuMsPerSearch": 0.3037, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:08:40.485Z to 2026-10-06T14:09:40.485Z, 240 samples, divided by the pass\u0027s 47151 completed searches", "waitedSeconds": 1, "busyBox": true, "busyReason": "processes outside the benchmark used 0.31 CPUs on average during the pass, above the limit of 0.3" }, { "target": "mongodb", "pass": "default@8", "startUtc": "2026-10-06T14:10:51.547Z", "endUtc": "2026-10-06T14:11:11.563Z", "seconds": 20.016, "endedBy": "result line", "governor": "performance", "frequencySamples": 77, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3491, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3495 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3495 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3503 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:10:36.545Z", "quietCheckLeadSeconds": 15.002, "outsideWindowFromUtc": "2026-10-06T14:09:36.545Z", "outsideLoadAtStart": 0.29, "outsideLoadRecentAtStart": 0.21, "outsideLoadDuring": 0.228, "accountingMismatch": false, "searches": 41327, "clientCpuSeconds": 11.8796, "clientCpuMsPerSearch": 0.2875, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:10:51.547Z to 2026-10-06T14:11:11.550Z, 77 samples, divided by the pass\u0027s 41327 completed searches", "waitedSeconds": 5, "busyBox": false }, { "target": "chroma", "pass": "default@8", "startUtc": "2026-10-06T14:13:38.370Z", "endUtc": "2026-10-06T14:13:58.383Z", "seconds": 20.013, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3487, "median": 3492, "max": 3495 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3497 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:13:20.353Z", "quietCheckLeadSeconds": 18.017, "outsideWindowFromUtc": "2026-10-06T14:12:20.353Z", "outsideLoadAtStart": 0.156, "outsideLoadRecentAtStart": 0.179, "outsideLoadDuring": 0.147, "accountingMismatch": false, "searches": 15547, "clientCpuSeconds": 7.4681, "clientCpuMsPerSearch": 0.4804, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:13:38.370Z to 2026-10-06T14:13:58.378Z, 80 samples, divided by the pass\u0027s 15547 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "chroma", "pass": "default@1", "startUtc": "2026-10-06T14:14:16.387Z", "endUtc": "2026-10-06T14:14:36.397Z", "seconds": 20.011, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:13:58.383Z", "quietCheckLeadSeconds": 18.003, "outsideWindowFromUtc": "2026-10-06T14:12:58.383Z", "outsideLoadAtStart": 0.157, "outsideLoadRecentAtStart": 0.137, "outsideLoadDuring": 0.159, "accountingMismatch": false, "searches": 9379, "clientCpuSeconds": 4.5046, "clientCpuMsPerSearch": 0.4803, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:14:16.387Z to 2026-10-06T14:14:36.388Z, 80 samples, divided by the pass\u0027s 9379 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "typesense", "pass": "default@8", "startUtc": "2026-10-06T14:16:37.737Z", "endUtc": "2026-10-06T14:16:57.746Z", "seconds": 20.009, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3489, "median": 3492, "max": 3493 }, { "cpu": 1, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3490, "median": 3492, "max": 3498 }, { "cpu": 5, "min": 3489, "median": 3492, "max": 3499 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:16:19.723Z", "quietCheckLeadSeconds": 18.014, "outsideWindowFromUtc": "2026-10-06T14:15:19.723Z", "outsideLoadAtStart": 0.211, "outsideLoadRecentAtStart": 0.193, "outsideLoadDuring": 0.041, "accountingMismatch": false, "searches": 11503, "clientCpuSeconds": 6.808, "clientCpuMsPerSearch": 0.5918, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:16:37.737Z to 2026-10-06T14:16:57.743Z, 80 samples, divided by the pass\u0027s 11503 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "typesense", "pass": "default@1", "startUtc": "2026-10-06T14:17:15.753Z", "endUtc": "2026-10-06T14:17:35.759Z", "seconds": 20.006, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3487, "median": 3492, "max": 3494 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 2, "min": 3491, "median": 3492, "max": 3497 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3497 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3486, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:16:57.748Z", "quietCheckLeadSeconds": 18.006, "outsideWindowFromUtc": "2026-10-06T14:15:57.748Z", "outsideLoadAtStart": 0.113, "outsideLoadRecentAtStart": 0.031, "outsideLoadDuring": 0.219, "accountingMismatch": false, "searches": 4705, "clientCpuSeconds": 2.4427, "clientCpuMsPerSearch": 0.5192, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:17:15.753Z to 2026-10-06T14:17:35.754Z, 80 samples, divided by the pass\u0027s 4705 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "typesense", "pass": "exact", "startUtc": "2026-10-06T14:17:53.768Z", "endUtc": "2026-10-06T14:18:53.780Z", "seconds": 60.012, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3486, "median": 3492, "max": 3500 }, { "cpu": 1, "min": 3399, "median": 3492, "max": 3502 }, { "cpu": 2, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3505 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3503 }, { "cpu": 6, "min": 3487, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3486, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:17:35.759Z", "quietCheckLeadSeconds": 18.009, "outsideWindowFromUtc": "2026-10-06T14:16:35.759Z", "outsideLoadAtStart": 0.15, "outsideLoadRecentAtStart": 0.216, "outsideLoadDuring": 0.209, "accountingMismatch": false, "searches": 10477, "clientCpuSeconds": 5.6648, "clientCpuMsPerSearch": 0.5407, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:17:53.768Z to 2026-10-06T14:18:53.772Z, 240 samples, divided by the pass\u0027s 10477 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sql-diskann", "pass": "default@1", "startUtc": "2026-10-06T14:20:57.685Z", "endUtc": "2026-10-06T14:21:17.693Z", "seconds": 20.009, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3498 }, { "cpu": 1, "min": 3490, "median": 3492, "max": 3498 }, { "cpu": 2, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3500 }, { "cpu": 6, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3490, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:20:39.683Z", "quietCheckLeadSeconds": 18.002, "outsideWindowFromUtc": "2026-10-06T14:19:39.683Z", "outsideLoadAtStart": 0.205, "outsideLoadRecentAtStart": 0.22, "outsideLoadDuring": 0.214, "accountingMismatch": false, "searches": 5430, "clientCpuSeconds": 5.5544, "clientCpuMsPerSearch": 1.0229, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:20:57.685Z to 2026-10-06T14:21:17.688Z, 80 samples, divided by the pass\u0027s 5430 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sql-diskann", "pass": "default@8", "startUtc": "2026-10-06T14:21:35.716Z", "endUtc": "2026-10-06T14:21:55.728Z", "seconds": 20.013, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3490, "median": 3492, "max": 3496 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3499 }, { "cpu": 2, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 3, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3499 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3495 }, { "cpu": 6, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 7, "min": 3492, "median": 3492, "max": 3492 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:21:17.695Z", "quietCheckLeadSeconds": 18.021, "outsideWindowFromUtc": "2026-10-06T14:20:17.695Z", "outsideLoadAtStart": 0.213, "outsideLoadRecentAtStart": 0.207, "outsideLoadDuring": 0.182, "accountingMismatch": false, "searches": 13820, "clientCpuSeconds": 15.1276, "clientCpuMsPerSearch": 1.0946, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:21:35.716Z to 2026-10-06T14:21:55.725Z, 80 samples, divided by the pass\u0027s 13820 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "sql-diskann", "pass": "exact", "startUtc": "2026-10-06T14:22:13.735Z", "endUtc": "2026-10-06T14:23:13.749Z", "seconds": 60.014, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3488, "median": 3492, "max": 3502 }, { "cpu": 1, "min": 3485, "median": 3492, "max": 3501 }, { "cpu": 2, "min": 3483, "median": 3492, "max": 3500 }, { "cpu": 3, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3486, "median": 3492, "max": 3501 }, { "cpu": 5, "min": 3486, "median": 3492, "max": 3504 }, { "cpu": 6, "min": 3488, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3489, "median": 3492, "max": 3501 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:21:55.729Z", "quietCheckLeadSeconds": 18.005, "outsideWindowFromUtc": "2026-10-06T14:20:55.729Z", "outsideLoadAtStart": 0.219, "outsideLoadRecentAtStart": 0.177, "outsideLoadDuring": 0.213, "accountingMismatch": false, "searches": 15660, "clientCpuSeconds": 18.1867, "clientCpuMsPerSearch": 1.1613, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:22:13.735Z to 2026-10-06T14:23:13.735Z, 240 samples, divided by the pass\u0027s 15660 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "duckdb", "pass": "default@8", "startUtc": "2026-10-06T14:25:04.124Z", "endUtc": "2026-10-06T14:25:24.132Z", "seconds": 20.008, "endedBy": "result line", "governor": "performance", "frequencySamples": 78, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 1, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 2, "min": 3490, "median": 3492, "max": 3503 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 5, "min": 3492, "median": 3492, "max": 3492 }, { "cpu": 6, "min": 3410, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3469, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:24:46.109Z", "quietCheckLeadSeconds": 18.015, "outsideWindowFromUtc": "2026-10-06T14:23:46.109Z", "outsideLoadAtStart": 0.209, "outsideLoadRecentAtStart": 0.183, "outsideLoadDuring": 0.206, "accountingMismatch": false, "searches": 11525, "clientCpuSeconds": 79.6373, "clientCpuMsPerSearch": 6.91, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:25:04.124Z to 2026-10-06T14:25:24.130Z, 78 samples, divided by the pass\u0027s 11525 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "duckdb", "pass": "exact", "startUtc": "2026-10-06T14:25:42.144Z", "endUtc": "2026-10-06T14:26:42.159Z", "seconds": 60.016, "endedBy": "result line", "governor": "performance", "frequencySamples": 240, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3489, "median": 3492, "max": 3502 }, { "cpu": 1, "min": 3486, "median": 3492, "max": 3507 }, { "cpu": 2, "min": 3485, "median": 3492, "max": 3503 }, { "cpu": 3, "min": 3489, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3486, "median": 3492, "max": 3497 }, { "cpu": 5, "min": 3490, "median": 3492, "max": 3498 }, { "cpu": 6, "min": 3489, "median": 3492, "max": 3503 }, { "cpu": 7, "min": 3489, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:25:24.135Z", "quietCheckLeadSeconds": 18.009, "outsideWindowFromUtc": "2026-10-06T14:24:24.135Z", "outsideLoadAtStart": 0.203, "outsideLoadRecentAtStart": 0.201, "outsideLoadDuring": 0.224, "accountingMismatch": false, "searches": 13956, "clientCpuSeconds": 87.4831, "clientCpuMsPerSearch": 6.2685, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:25:42.144Z to 2026-10-06T14:26:42.148Z, 240 samples, divided by the pass\u0027s 13956 completed searches", "waitedSeconds": 0, "busyBox": false }, { "target": "duckdb", "pass": "default@1", "startUtc": "2026-10-06T14:27:00.165Z", "endUtc": "2026-10-06T14:27:20.172Z", "seconds": 20.006, "endedBy": "result line", "governor": "performance", "frequencySamples": 80, "nearestSample": false, "engineMhzMedian": 3492, "clientMhzMedian": 3492, "cpuMhz": [ { "cpu": 0, "min": 3489, "median": 3492, "max": 3495 }, { "cpu": 1, "min": 3491, "median": 3492, "max": 3494 }, { "cpu": 2, "min": 3489, "median": 3492, "max": 3501 }, { "cpu": 3, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 4, "min": 3489, "median": 3492, "max": 3496 }, { "cpu": 5, "min": 3489, "median": 3492, "max": 3499 }, { "cpu": 6, "min": 3491, "median": 3492, "max": 3500 }, { "cpu": 7, "min": 3490, "median": 3492, "max": 3500 } ], "quietCheckBefore": "warm-up", "quietCheckUtc": "2026-10-06T14:26:42.160Z", "quietCheckLeadSeconds": 18.005, "outsideWindowFromUtc": "2026-10-06T14:25:42.160Z", "outsideLoadAtStart": 0.224, "outsideLoadRecentAtStart": 0.213, "outsideLoadDuring": 0.184, "accountingMismatch": false, "searches": 5393, "clientCpuSeconds": 27.0904, "clientCpuMsPerSearch": 5.0232, "clientCpuBasis": "user \u002B system CPU time of this process\u0027s threads (not its finished children; includes the 4 Hz sampler thread and the kernel network time spent on this process\u0027s behalf) from 2026-10-06T14:27:00.165Z to 2026-10-06T14:27:20.167Z, 80 samples, divided by the pass\u0027s 5393 completed searches", "waitedSeconds": 0, "busyBox": false } ], "cpuIdle": { "driver": "intel_idle", "governor": "menu", "intelIdleMaxCstate": "9", "states": [ { "index": 0, "name": "POLL", "latencyUs": 0, "residencyUs": 0, "disabledOn": "", "defaultStatus": "enabled" }, { "index": 1, "name": "C1", "latencyUs": 2, "residencyUs": 2, "disabledOn": "", "defaultStatus": "enabled" }, { "index": 2, "name": "C1E", "latencyUs": 10, "residencyUs": 20, "disabledOn": "", "defaultStatus": "enabled" }, { "index": 3, "name": "C3", "latencyUs": 33, "residencyUs": 100, "disabledOn": "0-7", "defaultStatus": "disabled" }, { "index": 4, "name": "C6", "latencyUs": 133, "residencyUs": 400, "disabledOn": "", "defaultStatus": "enabled" } ] }, "cpuIdleAtEnd": { "driver": "intel_idle", "governor": "menu", "intelIdleMaxCstate": "9", "states": [ { "index": 0, "name": "POLL", "latencyUs": 0, "residencyUs": 0, "disabledOn": "", "defaultStatus": "enabled" }, { "index": 1, "name": "C1", "latencyUs": 2, "residencyUs": 2, "disabledOn": "", "defaultStatus": "enabled" }, { "index": 2, "name": "C1E", "latencyUs": 10, "residencyUs": 20, "disabledOn": "", "defaultStatus": "enabled" }, { "index": 3, "name": "C3", "latencyUs": 33, "residencyUs": 100, "disabledOn": "0-7", "defaultStatus": "disabled" }, { "index": 4, "name": "C6", "latencyUs": 133, "residencyUs": 400, "disabledOn": "", "defaultStatus": "enabled" } ] }, "connections": [ { "target": "clickhouse", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-clickhouse", "address": "container-ip", "port": 8123, "containerId": "32cf2c26b272", "network": "gvb-clickhouse_default", "notUsed": "localhost:8123" } ], "observed": [ { "address": "localhost", "port": 1433, "sockets": 2, "kind": "not this target" }, { "address": "container-ip", "port": 8123, "sockets": 1, "kind": "this target" } ], "observedUtc": "2026-10-06T13:11:31.977Z" }, { "target": "vespa", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-vespa", "address": "container-ip", "port": 8080, "containerId": "d4d665486ac3", "network": "engines_default", "notUsed": "localhost:8090" }, { "what": "container gvb-vespa", "address": "container-ip", "port": 19071, "containerId": "d4d665486ac3", "network": "engines_default", "notUsed": "localhost:19071" } ], "observed": [ { "address": "container-ip", "port": 8080, "sockets": 1, "kind": "this target" } ], "observedUtc": "2026-10-06T13:17:28.263Z" }, { "target": "oracle", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-oracle", "address": "container-ip", "port": 1521, "containerId": "0da3dc27bd31", "network": "gvb-oracle_default", "notUsed": "localhost:1521" } ], "observed": [ { "address": "container-ip", "port": 1521, "sockets": 13, "kind": "this target" } ], "observedUtc": "2026-10-06T13:22:48.733Z" }, { "target": "elasticsearch", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-elasticsearch", "address": "container-ip", "port": 9200, "containerId": "f2716e40a93c", "network": "engines_default", "notUsed": "localhost:9200" } ], "observed": [ { "address": "container-ip", "port": 9200, "sockets": 8, "kind": "this target" } ], "observedUtc": "2026-10-06T13:27:31.248Z" }, { "target": "sql", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-mssql", "address": "container-ip", "port": 1433, "containerId": "5e402e3af1a6", "network": "gvb-mssql_default", "notUsed": "localhost:14330" } ], "observed": [ { "address": "container-ip", "port": 1433, "sockets": 9, "kind": "this target" } ], "observedUtc": "2026-10-06T13:30:02.922Z" }, { "target": "pgvector", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-pgvector", "address": "container-ip", "port": 5432, "containerId": "6d6073de90e0", "network": "gvb-pgvector_default", "notUsed": "localhost:5432" } ], "observed": [ { "address": "container-ip", "port": 5432, "sockets": 8, "kind": "this target" } ], "observedUtc": "2026-10-06T13:34:16.527Z" }, { "target": "sqlitevec", "hosting": "embedded", "route": "in this process (embedded), no network", "endpoints": [], "observed": [], "observedUtc": "2026-10-06T13:38:22.364Z", "problem": "no open TCP connection was found" }, { "target": "qdrant", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-qdrant", "address": "container-ip", "port": 6334, "containerId": "880915fef269", "network": "gvb-qdrant_default", "notUsed": "localhost:16334" }, { "what": "container gvb-qdrant", "address": "container-ip", "port": 6333, "containerId": "880915fef269", "network": "gvb-qdrant_default", "notUsed": "localhost:16333" } ], "observed": [ { "address": "container-ip", "port": 6334, "sockets": 1, "kind": "this target" } ], "observedUtc": "2026-10-06T13:40:46.214Z" }, { "target": "opensearch", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-opensearch", "address": "container-ip", "port": 9200, "containerId": "53b5350e8b2f", "network": "engines_default", "notUsed": "localhost:9201" } ], "observed": [ { "address": "container-ip", "port": 9200, "sockets": 8, "kind": "this target" } ], "observedUtc": "2026-10-06T13:45:21.896Z" }, { "target": "qdrant-hnsw", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-qdrant", "address": "container-ip", "port": 6334, "containerId": "69832df26c6f", "network": "gvb-qdrant_default", "notUsed": "localhost:16334" }, { "what": "container gvb-qdrant", "address": "container-ip", "port": 6333, "containerId": "69832df26c6f", "network": "gvb-qdrant_default", "notUsed": "localhost:16333" } ], "observed": [ { "address": "container-ip", "port": 6334, "sockets": 1, "kind": "this target" } ], "observedUtc": "2026-10-06T13:49:36.453Z" }, { "target": "redis", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-redis", "address": "container-ip", "port": 6379, "containerId": "26d9ff0dafd6", "network": "gvb-redis_default", "notUsed": "localhost:6379" } ], "observed": [ { "address": "container-ip", "port": 6379, "sockets": 1, "kind": "this target" } ], "observedUtc": "2026-10-06T13:54:02.960Z" }, { "target": "milvus", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-milvus", "address": "container-ip", "port": 19530, "containerId": "b9f11218b6aa", "network": "gvb-milvus_default", "notUsed": "localhost:19530" }, { "what": "container gvb-milvus", "address": "container-ip", "port": 9091, "containerId": "b9f11218b6aa", "network": "gvb-milvus_default", "notUsed": "localhost:9091" } ], "observed": [ { "address": "container-ip", "port": 19530, "sockets": 8, "kind": "this target" } ], "observedUtc": "2026-10-06T13:58:49.785Z" }, { "target": "mariadb", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvbbench-mariadb", "address": "container-ip", "port": 3306, "containerId": "52cb67e69c35", "network": "gvbbench-mariadb_default", "notUsed": "localhost:13306" } ], "observed": [ { "address": "container-ip", "port": 3306, "sockets": 8, "kind": "this target" } ], "observedUtc": "2026-10-06T14:03:09.906Z" }, { "target": "weaviate", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-weaviate", "address": "container-ip", "port": 8080, "containerId": "41c8b7799a14", "network": "gvb-weaviate_default", "notUsed": "localhost:8085" } ], "observed": [ { "address": "container-ip", "port": 8080, "sockets": 8, "kind": "this target" } ], "observedUtc": "2026-10-06T14:05:49.970Z" }, { "target": "mongodb", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-mongodb", "address": "container-ip", "port": 27017, "containerId": "441920b747a8", "network": "gvb-mongodb_default", "notUsed": "localhost:27017" } ], "observed": [ { "address": "container-ip", "port": 27017, "sockets": 10, "kind": "this target" } ], "observedUtc": "2026-10-06T14:11:11.563Z" }, { "target": "chroma", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-chroma", "address": "container-ip", "port": 8000, "containerId": "94961ec20450", "network": "gvb-chroma_default", "notUsed": "localhost:8000" } ], "observed": [ { "address": "container-ip", "port": 8000, "sockets": 8, "kind": "this target" } ], "observedUtc": "2026-10-06T14:14:36.397Z" }, { "target": "typesense", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-typesense", "address": "container-ip", "port": 8108, "containerId": "77eb2dcbee94", "network": "engines_default", "notUsed": "localhost:8108" } ], "observed": [ { "address": "container-ip", "port": 8108, "sockets": 1, "kind": "this target" } ], "observedUtc": "2026-10-06T14:18:53.780Z" }, { "target": "sql-diskann", "hosting": "compose", "route": "container address on its Docker network, not the published port (no docker-proxy)", "endpoints": [ { "what": "container gvb-mssql", "address": "container-ip", "port": 1433, "containerId": "39d172c6c285", "network": "gvb-mssql_default", "notUsed": "localhost:14330" } ], "observed": [ { "address": "container-ip", "port": 1433, "sockets": 9, "kind": "this target" } ], "observedUtc": "2026-10-06T14:23:13.749Z" }, { "target": "duckdb", "hosting": "embedded", "route": "in this process (embedded), no network", "endpoints": [], "observed": [], "observedUtc": "2026-10-06T14:27:20.172Z", "problem": "no open TCP connection was found" } ], "cpuAccounting": "busy = user \u002B nice \u002B system \u002B irq \u002B softirq \u002B steal (guest time is already in user); CONFIG_IRQ_TIME_ACCOUNTING is not set, so task and cgroup run time include the interrupt and softirq time that hit them and it is added; CONFIG_PARAVIRT_TIME_ACCOUNTING is not set, so steal is added (/boot/config-6.8.0-142-generic)", "loadAverageEnd": "3.50 3.94 3.51", "stateFile": "~/gvb-work/bench-machine-state.json", "restored": [ "client affinity back to 0-7", "governor back to schedutil on CPUs 0-7", "CPU clock back: turbo on (/sys/devices/system/cpu/intel_pstate/no_turbo 0); every CPU\u0027s ceiling and floor read as before the run; uncore limit back to min 1200 MHz, max 3000 MHz (MSR 0x620 = 0xc1e)" ], "restoreProblems": [], "flags": [ "sqlitevec is embedded: it ran inside the client process on the client CPUs 0-1,4-5, sharing them with the client.", "duckdb is embedded: it ran inside the client process on the client CPUs 0-1,4-5, sharing them with the client.", "WARNING: busy box during mongodb exact: processes outside the benchmark used 0.31 CPUs on average during the pass, above the limit of 0.3.", "WARNING: clock off its pinned value during oracle default@8: the engine CPUs averaged 3121 MHz, -10.8% against the pinned 3500 MHz; the client CPUs averaged 3141 MHz, -10.3% against the pinned 3500 MHz (limit 1%); this pass is not comparable with passes at the pinned clock." ] } }