Engines

Generated page

Generated from the result rows and the override registry; do not edit by hand. Regenerated at every landing.

Generated on 2026-10-10. Row set: 1050 rows, sha256 df6b74c069034ff1ef74b57dc19be46d5ce6e67e088a0595c84efb5301b5dbb1 (of the sorted row ids, one per line: the clean rows of the selection and the failed cells listed below).

For every engine in every lane, this page states what we run: the version, the container image digest where a row records one, the durability classes measured, the tiers, every setting a row carries, and each override of an engine's default with the reason. Everything below is read from the result rows and the override registry, so it cannot drift from what was measured. The rules these settings follow are on the Methodology page.

The rows are those of the campaign with instrument 2026-10, selected as the published tables are: the newest clean row of each cell, on the serial CPU set. A failed attempt is listed as the result it is.

What we run

Tables

ArangoDB: arangodb_tpc

Tiers	Version	Image	Durability classes
tpch1, tpch10	arangodb:3.12.11	arangodb@sha256:563cb2c07af0aead37fd688b58f51d6eb534a3da6163621e130e67d7a55176c4	relaxed, strict

• Durability, relaxed: waitForSync=false (default); RocksDB WAL synced every 100 ms
• Durability, strict: waitForSync=true: the commit waits for the WAL sync
• Setting topology: client_server
• Setting mem_cap: 16g (tpch1); 32g (tpch10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (tpch1); 32g (tpch10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

DuckDB: duckdb

Tiers	Version	Image	Durability classes
tpch1, tpch10	1.5.4	not recorded in the rows	relaxed, strict

• Durability, relaxed: fsync at commit, not configurable (DuckDB WAL)
• Durability, strict: fsync at commit, not configurable (DuckDB WAL)
• Setting topology: embedded
• Setting mem_cap: 16g (tpch1); 32g (tpch10)
• Setting cpuset: 0-11

Overrides of the engine's default on this arm:

• DuckDB is given one thread for each CPU the run may use; left alone it sizes its pool from the machine's cores and not from the CPUs the run was given. Override duckdb_threads.

MongoDB: mongodb

Tiers	Version	Image	Durability classes
tpch1, tpch10	mongodb 8.2.12	mongo@sha256:41afd6e1183f57e4e4d03ab733070671fca8553da2b36f15d6e3fc9760494d17	relaxed, strict

• Durability, relaxed: write concern w=1, j=false (journal flushed every 100 ms)
• Durability, strict: write concern w=1, j=true (the journal is synced before the ack)
• Setting topology: client_server
• Setting mem_cap: 16g (tpch1); 32g (tpch10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (tpch1); 32g (tpch10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

PostgreSQL: postgres

Tiers	Version	Image	Durability classes
tpch1, tpch10	PostgreSQL 18.6	postgres@sha256:7341002d2b8c7c5bdd7542a671a95b36196c0b5b888daf454ae4fc33ba5346d7	relaxed, strict

• Durability, relaxed: synchronous_commit=off
• Durability, strict: synchronous_commit=on; server flags synchronous_commit=on
• Setting topology: client_server
• Setting mem_cap: 16g (tpch1); 32g (tpch10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (tpch1); 32g (tpch10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

PostgreSQL (tuned): postgres_tuned

This arm is an ablation of one engine's image defaults. It is measured and kept in the rows, and it is not printed in the result tables.

Tiers	Version	Image	Durability classes
tpch1, tpch10	PostgreSQL 18.6	postgres@sha256:7341002d2b8c7c5bdd7542a671a95b36196c0b5b888daf454ae4fc33ba5346d7	relaxed, strict

• Durability, relaxed: synchronous_commit=off
• Durability, strict: synchronous_commit=on; server flags synchronous_commit=on
• Setting topology: client_server
• Setting mem_cap: 16g (tpch1); 32g (tpch10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (tpch1); 32g (tpch10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

SQLite: sqlite

Tiers	Version	Image	Durability classes
tpch1, tpch10	sqlite 3.46.1	not recorded in the rows	relaxed, strict

• Durability, relaxed: WAL, synchronous=NORMAL: synced at checkpoint, not at commit
• Durability, strict: WAL, synchronous=FULL: synced at every commit
• Setting topology: embedded
• Setting mem_cap: 16g (tpch1); 32g (tpch10)
• Setting cpuset: 0-11

No override from the registry applies to this arm.

SurrealDB (embedded): surrealdb_tpc

Tiers	Version	Image	Durability classes
tpch1	surrealdb-embedded:2.3.10 (sdk 2.0.0)	not recorded in the rows	relaxed, strict

• Durability, relaxed: SurrealKV, SURREAL_SYNC_DATA unset (the default): no sync at commit
• Durability, strict: SurrealKV, SURREAL_SYNC_DATA=true: sync at commit
• Setting topology: embedded
• Setting mem_cap: 16g
• Setting cpuset: 0-11

No override from the registry applies to this arm.

Failed cells (a failure is a result):

• tpch10 / olap / relaxed / rep 1: rc 137, killed for memory: last line: PHASE build-running t=3369.2s rss=32345.8MiB
• tpch10 / oltp / relaxed / rep 1: rc 137, killed for memory: last line: PHASE build-running t=3370.7s rss=32500.2MiB
• tpch10 / oltp / strict / rep 1: rc 137, killed for memory: last line: PHASE build-running t=3371.8s rss=31952.5MiB

SurrealDB (server): surrealdb_tpc_server

Tiers	Version	Image	Durability classes
tpch1, tpch10	surrealdb-server:3.2.4+20260803.93ab219	surrealdb/surrealdb@sha256:6a5002363ff5b000b72a55f985203e951e3175e578002954b0e38f113e48a698	relaxed, strict

• Durability, relaxed: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified
• Durability, strict: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified
• Setting topology: client_server
• Setting mem_cap: 16g (tpch1); 32g (tpch10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (tpch1); 32g (tpch10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

Failed cells (a failure is a result):

• tpch10 / olap / relaxed / rep 1: rc 1: socket.gaierror: [Errno -3] Temporary failure in name resolution

Registered for this lane, no row yet: ArcadeDB (embedded) (arcadedb_embedded, lane l1tpc), ArcadeDB (server) (arcadedb_server, lane l1tpc), ArcadeDB (server, image JVM defaults) (arcadedb_imgdefaults_server, lane l1tpc).

Graphs

ArangoDB: arangodb_graph

Tiers	Version	Image	Durability classes
sf1, sf10	arangodb:3.12.11	arangodb@sha256:563cb2c07af0aead37fd688b58f51d6eb534a3da6163621e130e67d7a55176c4	relaxed, strict
sf1, sf10	arangodb:3.12.12	arangodb@sha256:4bc086d5050ca7ea11c6d00a36d8b910c838bb54ad553f8c1b715769d3499bcf	relaxed, strict

• Durability, relaxed: waitForSync=false (default); RocksDB WAL synced every 100 ms
• Durability, strict: waitForSync=true: the commit waits for the WAL sync
• Setting topology: client_server
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting query_language: AQL

No override from the registry applies to this arm.

ArcadeDB (embedded): arcadedb_graph_embedded

Tiers	Version	Image	Durability classes
sf1, sf10, sf1full	26.10.1, commit d36b4ca3a	not recorded in the rows	relaxed, strict

• Durability, relaxed: txWalFlush=0 (engine default): no flush at commit
• Durability, strict: txWalFlush=1: the WAL is flushed with a data-only sync (fdatasync) at every commit
• Setting topology: embedded
• Setting mem_cap: 8g (sf1); 24g (sf10); 32g (sf1full)
• Setting heap: 4g (sf1); 12g (sf10); 16g (sf1full)
• Setting cpuset: 0-11
• Setting query_language: openCypher (the engine also has its own SQL)

No override from the registry applies to this arm.

ArcadeDB (server): arcadedb_graph_server

Tiers	Version	Image	Durability classes
sf1, sf10, sf1full	server:26.10.1 (build d36b4ca3ae4c170abc73598e0dffa2bb58e06621/1791229620009/main), commit d36b4ca3a	arcadedb-c25:26.10.1; digest sha256:8859153bae95b339a2c5453609fc760b283932e13be419a01055aaf451d221e5	relaxed, strict

• Durability, relaxed: txWalFlush=0 (engine default): no flush at commit (set on the server's JAVA_OPTS by runner.py and recorded as durability_server_flags; the server exposes no read-back)
• Durability, strict: txWalFlush=1: the WAL is flushed with a data-only sync (fdatasync) at every commit (set on the server's JAVA_OPTS by runner.py and recorded as durability_server_flags; the server exposes no read-back); server flags txWalFlush=1
• Setting topology: client_server
• Setting mem_cap: 8g (sf1); 24g (sf10); 32g (sf1full)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (sf1); 24g (sf10); 32g (sf1full)
• Setting heap: 4g (sf1); 12g (sf10); 16g (sf1full)
• Setting server_heap: 4g (sf1); 12g (sf10); 16g (sf1full)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting server_jvm_gc: G1
• Setting server_jvm_flags: -XX:HeapDumpPath=/home/arcadedb/log -XX:+UseCompactObjectHeaders -XX:+UseCompactObjectHeaders -Xms4g -Xmx4g -XX:+HeapDumpOnOutOfMemoryError (sf1); -XX:HeapDumpPath=/home/arcadedb/log -XX:+UseCompactObjectHeaders -XX:+UseCompactObjectHeaders -Xms12g -Xmx12g -XX:+HeapDumpOnOutOfMemoryError (sf10); -XX:HeapDumpPath=/home/arcadedb/log -XX:+UseCompactObjectHeaders -XX:+UseCompactObjectHeaders -Xms16g -Xmx16g -XX:+HeapDumpOnOutOfMemoryError (sf1full)
• Setting query_language: openCypher over HTTP (the engine also has its own SQL)
• Setting server_query_max_heap_elements: 5000000
• Setting server_query_max_heap_source: read from the engine over HTTP (schema:database)
• Setting server_http_body_max_bytes: 68719476736
• Setting server_http_body_max_source: read from the engine over HTTP (server?mode=default)
• Setting server_http_body_max_default: 104857600

Overrides of the engine's default on this arm:

• The ArcadeDB server is started with the limit on the records or groups that one sorting, grouping, or distinct query may hold in memory fixed at 5,000,000, while the embedded package leaves that limit at the engine's default, which grows with the heap, so the two deployments can run different limits. Override arcadedb_query_cap. Recorded on the rows: server_query_max_heap_elements: 5000000; server_query_max_heap_source: read from the engine over HTTP (schema:database).
• The ArcadeDB server that loads its data through the bulk endpoint is started with the limit on one HTTP request body raised to 64 GiB, where the engine's default is 100 MiB. The default refuses the single streamed request that carries the larger graph and cross-model loads (the engine's own message says to raise the limit or split the payload), so the load stays one request, as the embedded arm's is one call. The limit caps how much one request may carry and is no speed setting; the embedded arms have no HTTP body. Override arcadedb_http_body_limit. Recorded on the rows: server_http_body_max_bytes: 68719476736; server_http_body_max_source: read from the engine over HTTP (server?mode=default); server_http_body_max_default: 104857600.

DuckPGQ: duckpgq_graph

Tiers	Version	Image	Durability classes
sf1, sf10	duckdb:1.5.4 + duckpgq:f386a6c	not recorded in the rows	relaxed, strict

• Durability, relaxed: fsync at commit, not configurable (DuckDB WAL)
• Durability, strict: fsync at commit, not configurable (DuckDB WAL)
• Setting topology: embedded
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting duckpgq_threads: 12

Overrides of the engine's default on this arm:

• DuckDB is given one thread for each CPU the run may use; left alone it sizes its pool from the machine's cores and not from the CPUs the run was given. Override duckdb_threads. Recorded on the rows: duckpgq_threads: 12.

Failed cells (a failure is a result):

• sf1 / oltp / relaxed / rep 1: rc 1: _duckdb.OutOfMemoryException: Out of Memory Error: failed to allocate data of size 8.0 GiB (684.3 MiB/6.3 GiB used)
• sf1 / oltp / strict / rep 1: rc 1: _duckdb.OutOfMemoryException: Out of Memory Error: failed to allocate data of size 8.0 GiB (684.3 MiB/6.3 GiB used)
• sf10 / oltp / relaxed / rep 1: rc 1: _duckdb.OutOfMemoryException: Out of Memory Error: failed to allocate data of size 32.0 GiB (5.0 GiB/19.1 GiB used)
• sf10 / oltp / strict / rep 1: rc 1: _duckdb.OutOfMemoryException: Out of Memory Error: failed to allocate data of size 32.0 GiB (5.0 GiB/19.1 GiB used)

FalkorDB: falkordb_graph

Tiers	Version	Image	Durability classes
sf1, sf10	falkordb:4.20.6 (redis 8.6.3)	falkordb/falkordb@sha256:0a9fe4d1ee0bdda8e0a85ff36d3f03ca9ab91cd9441c4a836afae32e4014e2f7	relaxed, strict
sf1, sf10	falkordb:6.0.1 (redis 8.10.2)	falkordb/falkordb@sha256:2756fdea96acff753e49c459dc279a21a50915d7b2b38bff8a3950d10f984fd8	relaxed, strict

• Durability, relaxed: appendonly=no, RDB save '3600 1 300 100 60 10000' (image default): nothing is synced at commit
• Durability, strict: appendonly=yes, appendfsync=always: the AOF is fdatasynced at every write; server flags appendonly=yes, appendfsync=always
• Setting topology: client_server
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting query_language: Cypher over the Redis protocol
• Setting falkordb_appendonly: no (sf1, sf10); yes (sf1, sf10)
• Setting falkordb_appendfsync: always (sf1, sf10); everysec (sf1, sf10)
• Setting falkordb_save: 3600 1 300 100 60 10000

No override from the registry applies to this arm.

LadybugDB: ladybug_graph

Tiers	Version	Image	Durability classes
sf1, sf10	ladybug:0.20.4	not recorded in the rows	relaxed, strict
sf1, sf10	ladybug:0.21.2	not recorded in the rows	relaxed, strict

• Durability, relaxed: fsync at commit, not configurable (LadybugDB WAL)
• Durability, strict: fsync at commit, not configurable (LadybugDB WAL)
• Setting topology: embedded
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting query_language: Cypher, embedded

No override from the registry applies to this arm.

Memgraph: memgraph_graph

Tiers	Version	Image	Durability classes
sf1, sf10	memgraph:3.13.1	memgraph/memgraph@sha256:4710bee1ab5b47599876e30f17ae1679d0bbb2262d84dc06641521fecb7c89ce	relaxed, strict

• Durability, relaxed: storage-wal-enabled=true, storage-wal-file-flush-every-n-tx=100000 (image default): the WAL is fsynced every 100,000 transactions, not at commit
• Durability, strict: storage-wal-file-flush-every-n-tx=1: the WAL is fsynced at every commit; server flags storage-wal-file-flush-every-n-tx=1
• Setting topology: client_server
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting query_language: Cypher over Bolt
• Setting memgraph_storage_mode: IN_MEMORY_TRANSACTIONAL
• Setting memgraph_wal_enabled: true

No override from the registry applies to this arm.

MongoDB: mongodb_graph

Tiers	Version	Image	Durability classes
sf1, sf10	mongodb 8.2.12	mongo@sha256:41afd6e1183f57e4e4d03ab733070671fca8553da2b36f15d6e3fc9760494d17	relaxed, strict

• Durability, relaxed: write concern w=1, j=false (journal flushed every 100 ms)
• Durability, strict: write concern w=1, j=true (the journal is synced before the ack)
• Setting topology: client_server
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting query_language: the aggregation pipeline ($lookup joins over both ends of each friendship)

No override from the registry applies to this arm.

Neo4j: neo4j_graph

Tiers	Version	Image	Durability classes
sf1, sf10	neo4j:2026.08.1	neo4j@sha256:e702d6b535d9d3ae01ee7b132ec87aa40e23d3f0ace82fbfc344e2048cb81960	relaxed, strict

• Durability, relaxed: fsync at commit, not configurable (no durability setting in SHOW SETTINGS at 2026.08.1)
• Durability, strict: fsync at commit, not configurable (no durability setting in SHOW SETTINGS at 2026.08.1)
• Setting topology: client_server
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (sf1); 24g (sf10)
• Setting heap: 4g (sf1); 12g (sf10)
• Setting server_heap: 4g (sf1); 12g (sf10)
• Setting server_pagecache: 3.0g (sf1); 11.0g (sf10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting query_language: Cypher over Bolt
• Setting neo4j_pagecache: 3.00GiB (sf1); 11.00GiB (sf10)
• Setting neo4j_checkpoint_interval: 5s
• Setting neo4j_checkpoint_interval_default: 15m

Overrides of the engine's default on this arm:

• Neo4j's page cache is sized from the memory the run was given, what is left after its heap and a fixed reserve, where its image would set one small fixed size whatever the container holds. Override neo4j_pagecache. Recorded on the rows: neo4j_pagecache: 3.00GiB (sf1); 11.00GiB (sf10).
• Neo4j checkpoints its store every 5 seconds, where its default is every 15 minutes, so that the disk reading taken after a run finds the data on disk. Override neo4j_checkpoint. Recorded on the rows: neo4j_checkpoint_interval: 5s; neo4j_checkpoint_interval_default: 15m.

SurrealDB (embedded): surrealdb_graph

Tiers	Version	Image	Durability classes
sf1, sf10	surrealdb-embedded:2.3.10 (sdk 2.0.0)	not recorded in the rows	relaxed, strict

• Durability, relaxed: SurrealKV, SURREAL_SYNC_DATA unset (the default): no sync at commit
• Durability, strict: SurrealKV, SURREAL_SYNC_DATA=true: sync at commit
• Setting topology: embedded
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting query_language: SurrealQL, embedded

No override from the registry applies to this arm.

SurrealDB (server): surrealdb_graph_server

Tiers	Version	Image	Durability classes
sf1, sf10	surrealdb-server:3.2.4+20260803.93ab219	surrealdb/surrealdb@sha256:6a5002363ff5b000b72a55f985203e951e3175e578002954b0e38f113e48a698	relaxed, strict

• Durability, relaxed: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified; RocksDB, sync=never: no sync at commit (the OS flushes)
• Durability, strict: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified; RocksDB, sync=every (the engine default): synced at every commit; server flags sync=every
• Setting topology: client_server
• Setting mem_cap: 8g (sf1); 24g (sf10)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (sf1); 24g (sf10)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting query_language: SurrealQL over WebSocket
• Setting surreal_sync_mode: every transaction commit (sf1, sf10); never (sf1, sf10)

No override from the registry applies to this arm.

Registered for this lane, no row yet: PostgreSQL + AGE (pgage_graph, lane l2).

Vectors

ArangoDB: arangodb_dense

Tiers	Version	Image	Durability classes
small	arangodb:3.12.11	arangodb@sha256:563cb2c07af0aead37fd688b58f51d6eb534a3da6163621e130e67d7a55176c4	relaxed

• Durability, relaxed: waitForSync=false (default); RocksDB WAL synced every 100 ms
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

Chroma: chroma_dense

Tiers	Version	Image	Durability classes
small, deep10m	1.5.9	dbbench:dense (a tag; no digest in the rows)	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: embedded
• Setting mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

DuckDB VSS: duckdb_vss_dense

Tiers	Version	Image	Durability classes
small, deep10m	1.5.4	dbbench:dense (a tag; no digest in the rows)	relaxed

• Durability, relaxed: fsync at commit, not configurable (DuckDB WAL)
• Setting topology: embedded
• Setting mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting quantization: fp32

Overrides of the engine's default on this arm:

• DuckDB is given one thread for each CPU the run may use; left alone it sizes its pool from the machine's cores and not from the CPUs the run was given. Override duckdb_threads.
• DuckDB's vector rows keep their HNSW index in the database through a persistence feature that its vector extension still labels experimental and that has to be switched on for the index to be stored at all. Override duckdb_vss.

LanceDB: lancedb_dense

Tiers	Version	Image	Durability classes
small, deep10m	0.39.0	dbbench:dense (a tag; no digest in the rows)	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: embedded
• Setting mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting quantization: INT8

No override from the registry applies to this arm.

Milvus: milvus_dense

Tiers	Version	Image	Durability classes
small, deep10m	milvus:3.0.1	milvusdb/milvus@sha256:2b2fc2cf499ad897c93d4b90ee251646f718522845261a342d74d7c0c66bb274	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

Milvus: milvus_dense_int8

Tiers	Version	Image	Durability classes
small, deep10m	milvus:3.0.1	milvusdb/milvus@sha256:2b2fc2cf499ad897c93d4b90ee251646f718522845261a342d74d7c0c66bb274	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: INT8

No override from the registry applies to this arm.

MongoDB: mongodb_dense

Tiers	Version	Image	Durability classes
small, deep10m	mongodb 8.2.12 + MongoDb Search Community Version 1.70.4 at localhost:27028	dbbench:mongo-search	relaxed

• Durability, relaxed: write concern w=1, j=false (journal flushed every 100 ms)
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

Neo4j: neo4j_dense

Tiers	Version	Image	Durability classes
small, deep10m	neo4j:2026.08.1	neo4j@sha256:e702d6b535d9d3ae01ee7b132ec87aa40e23d3f0ace82fbfc344e2048cb81960	relaxed

• Durability, relaxed: fsync at commit, not configurable (no durability setting in SHOW SETTINGS at 2026.08.1)
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (small); 36g (deep10m)
• Setting heap: 8g (small); 24g (deep10m)
• Setting server_heap: 8g (small); 24g (deep10m)
• Setting server_pagecache: 7.0g (small); 11.0g (deep10m)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: fp32

Overrides of the engine's default on this arm:

• Neo4j's page cache is sized from the memory the run was given, what is left after its heap and a fixed reserve, where its image would set one small fixed size whatever the container holds. Override neo4j_pagecache.

pgvector: pgvector_dense

Tiers	Version	Image	Durability classes
small, deep10m	pgvector:0.8.6 on PostgreSQL 18.6	pgvector/pgvector@sha256:1d50c689b0a6511b9ea0a15615281c81a59fd04a08eb35057ec8646fb3a2118a	relaxed

• Durability, relaxed: synchronous_commit=off
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

Qdrant: qdrant_dense

Tiers	Version	Image	Durability classes
small, deep10m	qdrant:1.19.1	qdrant/qdrant@sha256:0699e7733a6fa7fa7f6b95dcbed84ebb04584110da525cdfdef9f305c4f57738	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

Qdrant: qdrant_dense_int8

Tiers	Version	Image	Durability classes
small, deep10m	qdrant:1.19.1	qdrant/qdrant@sha256:0699e7733a6fa7fa7f6b95dcbed84ebb04584110da525cdfdef9f305c4f57738	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: INT8

No override from the registry applies to this arm.

sqlite-vec: sqlite_vec_dense

Tiers	Version	Image	Durability classes
small, deep10m	0.1.9	dbbench:dense (a tag; no digest in the rows)	relaxed

• Durability, relaxed: WAL, synchronous=NORMAL: synced at checkpoint, not at commit
• Setting topology: embedded
• Setting mem_cap: 16g (small); 36g (deep10m)
• Setting cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

sqlite-vec: sqlite_vec_dense_int8

Tiers	Version	Image	Durability classes
deep10m	0.1.9	dbbench:dense (a tag; no digest in the rows)	relaxed

• Durability, relaxed: WAL, synchronous=NORMAL: synced at checkpoint, not at commit
• Setting topology: embedded
• Setting mem_cap: 36g
• Setting cpuset: 0-11
• Setting quantization: INT8

No override from the registry applies to this arm.

SurrealDB (embedded): surrealdb_dense

No clean row yet; the failed cells below are all there is.

Failed cells (a failure is a result):

• small / search / relaxed / rep 1: rc -1: timeout_after_7200s
• deep10m / search / relaxed / rep 1: rc -1: timeout_after_28800s

SurrealDB (server): surrealdb_dense_server

Tiers	Version	Image	Durability classes
small	surrealdb-server:3.2.4+20260803.93ab219	surrealdb/surrealdb@sha256:6a5002363ff5b000b72a55f985203e951e3175e578002954b0e38f113e48a698	relaxed

• Durability, relaxed: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11
• Setting quantization: fp32

No override from the registry applies to this arm.

Failed cells (a failure is a result):

• deep10m / search / relaxed / rep 1: rc 1: socket.gaierror: [Errno -3] Temporary failure in name resolution

Elasticsearch: elasticsearch_sparse

Tiers	Version	Image	Durability classes
tiny, small, medium	9.5.4	docker.elastic.co/elasticsearch/elasticsearch@sha256:33178ff49e06da93e3c51c5d87401b26e7a6dea0ef9bb26539cfddc46478b420	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (tiny); 16g (small); 32g (medium)
• Setting heap: 4g (tiny); 8g (small); 16g (medium)
• Setting server_heap: 4g (tiny); 8g (small); 16g (medium)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

Overrides of the engine's default on this arm:

• Elasticsearch runs with its security features switched off, so none of its requests pays for authentication or encryption; its image turns both on unless told otherwise. Override es_security.
• Elasticsearch's index is created with no replica, because a single-node cluster has nowhere to place one and the index would otherwise stay yellow. Override es_replicas.

Milvus: milvus_sparse

Tiers	Version	Image	Durability classes
tiny, small, medium	milvus:3.0.1	milvusdb/milvus@sha256:2b2fc2cf499ad897c93d4b90ee251646f718522845261a342d74d7c0c66bb274	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (tiny); 16g (small); 32g (medium)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

pgvector: pgvector_sparse

Tiers	Version	Image	Durability classes
tiny, small, medium	pgvector:0.8.6 on PostgreSQL 18.6	pgvector/pgvector@sha256:1d50c689b0a6511b9ea0a15615281c81a59fd04a08eb35057ec8646fb3a2118a	relaxed

• Durability, relaxed: synchronous_commit=off
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (tiny); 16g (small); 32g (medium)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

Qdrant: qdrant_sparse

Tiers	Version	Image	Durability classes
tiny, small, medium	qdrant:1.19.1	qdrant/qdrant@sha256:0699e7733a6fa7fa7f6b95dcbed84ebb04584110da525cdfdef9f305c4f57738	relaxed

• Durability, relaxed: engine default; no transactional write timed on this lane
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 8g (tiny); 16g (small); 32g (medium)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

Registered for this lane, no row yet: ArangoDB (arangodb_dense_int8, lane l3d), ArcadeDB (embedded) (arcadedb_dense_embedded, lane l3d), ArcadeDB (embedded) (arcadedb_dense_embedded_int8, lane l3d), ArcadeDB (embedded, fp32) (arcadedb_sparse_embedded_fp32, lane l3s), ArcadeDB (embedded, int8) (arcadedb_sparse_embedded, lane l3s), ArcadeDB (embedded, no settle step) (arcadedb_sparse_embedded_nocompact, lane l3s), ArcadeDB (server) (arcadedb_dense_server, lane l3d), ArcadeDB (server) (arcadedb_dense_server_int8, lane l3d), ArcadeDB (server, fp32) (arcadedb_sparse_server_fp32, lane l3s), ArcadeDB (server, int8) (arcadedb_sparse_server, lane l3s), Elasticsearch (elasticsearch_dense, lane l3d), Elasticsearch (elasticsearch_dense_int8, lane l3d), FalkorDB (falkordb_dense, lane l3d), LadybugDB (ladybug_dense, lane l3d), LanceDB (lancedb_dense_fp32, lane l3d), Memgraph (memgraph_dense, lane l3d), Memgraph (memgraph_dense_int8, lane l3d), MongoDB (mongodb_dense_int8, lane l3d), Neo4j (neo4j_dense_int8, lane l3d), Qdrant (qdrant_sparse_uint8, lane l3s).

Time series

ArangoDB: arangodb_ts

Tiers	Version	Image	Durability classes
ts100, ts1000	arangodb:3.12.11 (from backend_version)	arangodb@sha256:563cb2c07af0aead37fd688b58f51d6eb534a3da6163621e130e67d7a55176c4	relaxed

• Durability, relaxed: waitForSync=false (default); RocksDB WAL synced every 100 ms
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (ts100); 32g (ts1000)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

DuckDB: duckdb

Tiers	Version	Image	Durability classes
ts100, ts1000	duckdb 1.5.4 (from backend_version)	dbbench:duckdb (a tag; no digest in the rows)	relaxed

• Durability, relaxed: fsync at commit, not configurable (DuckDB WAL)
• Setting topology: embedded
• Setting mem_cap: 16g (ts100); 32g (ts1000)
• Setting cpuset: 0-11

Overrides of the engine's default on this arm:

• DuckDB is given one thread for each CPU the run may use; left alone it sizes its pool from the machine's cores and not from the CPUs the run was given. Override duckdb_threads.

MongoDB: mongodb

Tiers	Version	Image	Durability classes
ts100, ts1000	mongodb 8.2.12 (from backend_version)	mongo@sha256:41afd6e1183f57e4e4d03ab733070671fca8553da2b36f15d6e3fc9760494d17	relaxed

• Durability, relaxed: write concern w=1, j=false (journal flushed every 100 ms)
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (ts100); 32g (ts1000)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

QuestDB: questdb

Tiers	Version	Image	Durability classes
ts100, ts1000	questdb Build Information: QuestDB 10.0.1, JDK 25.0.2, Commit Hash 7a391566ac827e0d8c269b91f2415c16b0a32a84 (from backend_version)	questdb/questdb@sha256:931af4156771ee2948ec2431988c3a93b257487d993f1b74771545c8dce5c23d	relaxed

• Durability, relaxed: cairo.commit.mode=nosync (default): no fsync at commit
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (ts100); 32g (ts1000)
• Setting heap: 8g (ts100); 16g (ts1000)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

SQLite: sqlite

Tiers	Version	Image	Durability classes
ts100, ts1000	sqlite 3.46.1 (from backend_version)	dbbench:client (a tag; no digest in the rows)	relaxed

• Durability, relaxed: WAL, synchronous=NORMAL: synced at checkpoint, not at commit
• Setting topology: embedded
• Setting mem_cap: 16g (ts100); 32g (ts1000)
• Setting cpuset: 0-11

No override from the registry applies to this arm.

SurrealDB (embedded): surrealdb_ts

Tiers	Version	Image	Durability classes
ts100	surrealdb-embedded:2.3.10 (sdk 2.0.0) (from backend_version)	dbbench:client (a tag; no digest in the rows)	relaxed

• Durability, relaxed: SurrealKV, SURREAL_SYNC_DATA unset (the default): no sync at commit
• Setting topology: embedded
• Setting mem_cap: 16g
• Setting cpuset: 0-11

No override from the registry applies to this arm.

Failed cells (a failure is a result):

• ts1000 / ingest / relaxed / rep 1: rc 137, killed for memory: last line: PHASE ingest-running t=1923.0s rss=32510.0MiB

SurrealDB (server): surrealdb_ts_server

Tiers	Version	Image	Durability classes
ts100, ts1000	surrealdb-server:3.2.4+20260803.93ab219 (from backend_version)	surrealdb/surrealdb@sha256:6a5002363ff5b000b72a55f985203e951e3175e578002954b0e38f113e48a698	relaxed

• Durability, relaxed: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (ts100); 32g (ts1000)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

TimescaleDB: timescaledb

Tiers	Version	Image	Durability classes
ts100, ts1000	timescaledb 2.30.1 on PostgreSQL 18.6 on x86_64-pc-linux-musl, compiled by gcc (from backend_version)	timescale/timescaledb@sha256:f7036933154c52dbc500f7b08ff8e28404a8ccabbf8ce3528142cde6ab253eef	relaxed

• Durability, relaxed: synchronous_commit=off
• Setting topology: client_server
• Setting mem_cap: 8g
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 16g (ts100); 32g (ts1000)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

Registered for this lane, no row yet: ArcadeDB (arcadedb_ts_doc, lane l4), ArcadeDB (arcadedb_ts_doc_server, lane l4), ArcadeDB (arcadedb_ts_native, lane l4), ArcadeDB (arcadedb_ts_native_server, lane l4), PostgreSQL (postgres_ts, lane l4).

Cross-model

ArangoDB: arangodb_e2

Tiers	Version	Image	Durability classes
e2, e2_500k	arangodb:3.12.11	arangodb@sha256:563cb2c07af0aead37fd688b58f51d6eb534a3da6163621e130e67d7a55176c4	relaxed, strict

• Durability, relaxed: waitForSync=false (default); RocksDB WAL synced every 100 ms
• Durability, strict: waitForSync=true: the commit waits for the WAL sync
• Setting topology: client_server
• Setting mem_cap: 12g (e2); 32g (e2_500k)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 12g (e2); 32g (e2_500k)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

MongoDB: mongodb_e2

Tiers	Version	Image	Durability classes
e2, e2_500k	mongodb 8.2.12 + MongoDb Search Community Version 1.70.4 at localhost:27028	dbbench:mongo-search; digest sha256:447135ddc03eed653fd613e2a9a56ac6623ad4c3877ae6b9f9e240aa8be97a72	strict
e2, e2_500k	mongodb 8.2.12 + MongoDb Search Community Version 1.70.4 at localhost:27028	dbbench:mongo-search; digest sha256:64bf722bc986a217881308c2ea6bca134c04f80fe86ec49d8f2caf8a511d6eec	relaxed

• Durability, relaxed: write concern w=1, j=false (journal flushed every 100 ms)
• Durability, strict: write concern w=1, j=true (the journal is synced before the ack)
• Setting topology: client_server
• Setting mem_cap: 12g (e2); 32g (e2_500k)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 12g (e2); 32g (e2_500k)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

Neo4j (vector index): neo4j_e2

Tiers	Version	Image	Durability classes
e2, e2_500k	neo4j:2026.08.1	neo4j@sha256:e702d6b535d9d3ae01ee7b132ec87aa40e23d3f0ace82fbfc344e2048cb81960	relaxed, strict

• Durability, relaxed: fsync at commit, not configurable (no durability setting in SHOW SETTINGS at 2026.08.1)
• Durability, strict: fsync at commit, not configurable (no durability setting in SHOW SETTINGS at 2026.08.1)
• Setting topology: client_server
• Setting mem_cap: 12g (e2); 32g (e2_500k)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 12g (e2); 32g (e2_500k)
• Setting heap: 6g (e2); 16g (e2_500k)
• Setting server_heap: 6g (e2); 16g (e2_500k)
• Setting server_pagecache: 5.0g (e2); 15.0g (e2_500k)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

Overrides of the engine's default on this arm:

• Neo4j's page cache is sized from the memory the run was given, what is left after its heap and a fixed reserve, where its image would set one small fixed size whatever the container holds. Override neo4j_pagecache.

PostgreSQL + pgvector + AGE: pg_age_e2

Tiers	Version	Image	Durability classes
e2, e2_500k	PostgreSQL 18.6 + pgvector:0.8.6 + age:1.8.0	dbbench:pg-age; digest sha256:1ab18fe1d6c24dd0bf8022f3ac3ecf9432a153a184065701b332732e1aeba3bd	relaxed
e2, e2_500k	PostgreSQL 18.6 + pgvector:0.8.6 + age:1.8.0	dbbench:pg-age; digest sha256:cbcf2d6975b79bbae4564ea62ee0a28396fb666276123a9e0abc24e7044f9843	strict

• Durability, relaxed: synchronous_commit=off
• Durability, strict: synchronous_commit=on; server flags synchronous_commit=on
• Setting topology: client_server
• Setting mem_cap: 12g (e2); 32g (e2_500k)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 12g (e2); 32g (e2_500k)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

Qdrant + Neo4j (no shared transaction): composed_qdrant_neo4j

Tiers	Version	Image	Durability classes
e2, e2_500k	qdrant-local:1.19.1+neo4j:2026.08.1	neo4j@sha256:e702d6b535d9d3ae01ee7b132ec87aa40e23d3f0ace82fbfc344e2048cb81960	relaxed, strict

• Durability, relaxed: fsync at commit, not configurable (no durability setting in SHOW SETTINGS at 2026.08.1); Qdrant WAL at its default
• Durability, strict: fsync at commit, not configurable (no durability setting in SHOW SETTINGS at 2026.08.1); Qdrant WAL at its default
• Setting topology: client_server
• Setting mem_cap: 12g (e2); 32g (e2_500k)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 12g (e2); 32g (e2_500k)
• Setting heap: 6g (e2); 16g (e2_500k)
• Setting server_heap: 6g (e2); 16g (e2_500k)
• Setting server_pagecache: 5.0g (e2); 15.0g (e2_500k)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

Overrides of the engine's default on this arm:

• Neo4j's page cache is sized from the memory the run was given, what is left after its heap and a fixed reserve, where its image would set one small fixed size whatever the container holds. Override neo4j_pagecache.
• Neo4j checkpoints its store at a short fixed interval instead of its much longer default, so that the disk reading taken after a run finds the data on disk. Override neo4j_checkpoint.

SurrealDB (embedded): surrealdb_e2

Tiers	Version	Image	Durability classes
e2	surrealdb-embedded:2.3.10 (sdk 2.0.0)	not recorded in the rows	relaxed, strict

• Durability, relaxed: SurrealKV, SURREAL_SYNC_DATA unset (the default): no sync at commit
• Durability, strict: SurrealKV, SURREAL_SYNC_DATA=true: sync at commit
• Setting topology: embedded
• Setting mem_cap: 12g
• Setting cpuset: 0-11

No override from the registry applies to this arm.

Failed cells (a failure is a result):

• e2_500k / atomicity / relaxed / rep 1: rc -1: timeout_after_7200s
• e2_500k / hybrid / relaxed / rep 1: rc -1: timeout_after_7200s
• e2_500k / hybrid / strict / rep 1: rc -1: timeout_after_7200s

SurrealDB (server): surrealdb_e2_server

Tiers	Version	Image	Durability classes
e2, e2_500k	surrealdb-server:3.2.4+20260803.93ab219	surrealdb/surrealdb@sha256:6a5002363ff5b000b72a55f985203e951e3175e578002954b0e38f113e48a698	relaxed, strict

• Durability, relaxed: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified
• Durability, strict: RocksDB at the engine default; SurrealDB 3.2.4 exposes no sync setting and the behaviour at commit is not verified
• Setting topology: client_server
• Setting mem_cap: 12g (e2); 32g (e2_500k)
• Setting client_mem_cap: 8g
• Setting server_mem_cap: 12g (e2); 32g (e2_500k)
• Setting cpuset: 0-11
• Setting server_cpuset: 0-11

No override from the registry applies to this arm.

Registered for this lane, no row yet: ArcadeDB (one transaction) (arcadedb_e2, lane e2), ArcadeDB (server, one transaction) (arcadedb_e2_server, lane e2), DuckDB (vss + DuckPGQ) (duckdb_e2, lane e2), LadybugDB (vector extension) (ladybug_e2, lane e2), Memgraph (vector index) (memgraph_e2, lane e2).

Lifecycle

SurrealDB (embedded): surrealdb_lifecycle

Tiers	Version	Image	Durability classes
lc10k, lc100k, lc1m, lc10m	surrealdb-embedded:2.3.10 (sdk 2.0.0)	dbbench:client (a tag; no digest in the rows)	relaxed

• Durability, relaxed: SurrealKV, SURREAL_SYNC_DATA unset (the default): no sync at commit
• Setting topology: embedded
• Setting mem_cap: 8g (lc10k, lc100k); 24g (lc1m); 28g (lc10m)
• Setting cpuset: 0-11
• Setting store: surrealkv

No override from the registry applies to this arm.

Failed cells (a failure is a result):

• lc1m / vector / relaxed / rep 1: rc -1: timeout_after_28800s
• lc10m / doc_idx10 / relaxed / rep 1: rc 137, killed for memory: container exited 137 with no output (cgroup OOM kill: raise the envelope or bound the harness's peak memory)
• lc10m / graph / relaxed / rep 1: rc 137, killed for memory: container exited 137 with no output (cgroup OOM kill: raise the envelope or bound the harness's peak memory)
• lc10m / vector / relaxed / rep 1: rc 137, killed for memory: container exited 137 with no output (cgroup OOM kill: raise the envelope or bound the harness's peak memory)

Registered for this lane, no row yet: ArcadeDB (arcadedb_e4, lane e4), ArcadeDB (embedded) (arcadedb_embedded, lane lifecycle), ArcadeDB (server) (arcadedb_server, lane lifecycle).

Gaps in the rows

What the rows do not yet say, found by reading them. An empty list would mean the rows are complete.

• Every lane has at least one row.
• Arms registered in the runner with no row yet:
  • lane e2: arcadedb_e2
  • lane e2: arcadedb_e2_server
  • lane e2: duckdb_e2
  • lane e2: ladybug_e2
  • lane e2: memgraph_e2
  • lane e4: arcadedb_e4
  • lane l1tpc: arcadedb_embedded
  • lane l1tpc: arcadedb_imgdefaults_server
  • lane l1tpc: arcadedb_server
  • lane l2: pgage_graph
  • lane l3d: arangodb_dense_int8
  • lane l3d: arcadedb_dense_embedded
  • lane l3d: arcadedb_dense_embedded_int8
  • lane l3d: arcadedb_dense_server
  • lane l3d: arcadedb_dense_server_int8
  • lane l3d: elasticsearch_dense
  • lane l3d: elasticsearch_dense_int8
  • lane l3d: falkordb_dense
  • lane l3d: ladybug_dense
  • lane l3d: lancedb_dense_fp32
  • lane l3d: memgraph_dense
  • lane l3d: memgraph_dense_int8
  • lane l3d: mongodb_dense_int8
  • lane l3d: neo4j_dense_int8
  • lane l3s: arcadedb_sparse_embedded
  • lane l3s: arcadedb_sparse_embedded_fp32
  • lane l3s: arcadedb_sparse_embedded_nocompact
  • lane l3s: arcadedb_sparse_server
  • lane l3s: arcadedb_sparse_server_fp32
  • lane l3s: qdrant_sparse_uint8
  • lane l4: arcadedb_ts_doc
  • lane l4: arcadedb_ts_doc_server
  • lane l4: arcadedb_ts_native
  • lane l4: arcadedb_ts_native_server
  • lane l4: postgres_ts
  • lane lifecycle: arcadedb_embedded
  • lane lifecycle: arcadedb_server
• Arms with a recorded failure and no clean row:
  • SurrealDB (embedded) (surrealdb_dense), lane l3d
• Override stamps that at least one row lacks or contradicts:
  • DuckDB (duckdb), lane l1tpc: duckdb_threads for duckdb_threads is missing, tier tpch1
  • DuckDB (duckdb), lane l1tpc: duckdb_threads for duckdb_threads is missing, tier tpch10
  • DuckDB (duckdb), lane l4: duckdb_threads for duckdb_threads is missing, tier ts100
  • DuckDB (duckdb), lane l4: duckdb_threads for duckdb_threads is missing, tier ts1000
  • DuckDB VSS (duckdb_vss_dense), lane l3d: duckdb_hnsw_persistence for duckdb_vss is missing, tier deep10m
  • DuckDB VSS (duckdb_vss_dense), lane l3d: duckdb_hnsw_persistence for duckdb_vss is missing, tier small
  • DuckDB VSS (duckdb_vss_dense), lane l3d: duckdb_threads for duckdb_threads is missing, tier deep10m
  • DuckDB VSS (duckdb_vss_dense), lane l3d: duckdb_threads for duckdb_threads is missing, tier small
  • Elasticsearch (elasticsearch_sparse), lane l3s: es_replicas for es_replicas is missing, tier medium
  • Elasticsearch (elasticsearch_sparse), lane l3s: es_replicas for es_replicas is missing, tier small
  • Elasticsearch (elasticsearch_sparse), lane l3s: es_replicas for es_replicas is missing, tier tiny
  • Elasticsearch (elasticsearch_sparse), lane l3s: es_security_enabled for es_security is missing, tier medium
  • Elasticsearch (elasticsearch_sparse), lane l3s: es_security_enabled for es_security is missing, tier small
  • Elasticsearch (elasticsearch_sparse), lane l3s: es_security_enabled for es_security is missing, tier tiny
  • Neo4j (neo4j_dense), lane l3d: neo4j_pagecache for neo4j_pagecache is missing, tier deep10m
  • Neo4j (neo4j_dense), lane l3d: neo4j_pagecache for neo4j_pagecache is missing, tier small
  • Neo4j (neo4j_graph), lane l2: neo4j_checkpoint_interval for neo4j_checkpoint is missing, tier sf1
  • Neo4j (neo4j_graph), lane l2: neo4j_checkpoint_interval for neo4j_checkpoint is missing, tier sf10
  • Neo4j (neo4j_graph), lane l2: neo4j_pagecache for neo4j_pagecache is missing, tier sf1
  • Neo4j (neo4j_graph), lane l2: neo4j_pagecache for neo4j_pagecache is missing, tier sf10
  • Neo4j (vector index) (neo4j_e2), lane e2: neo4j_pagecache for neo4j_pagecache is missing, tier e2
  • Neo4j (vector index) (neo4j_e2), lane e2: neo4j_pagecache for neo4j_pagecache is missing, tier e2_500k
  • Qdrant + Neo4j (no shared transaction) (composed_qdrant_neo4j), lane e2: neo4j_checkpoint_interval for neo4j_checkpoint is missing, tier e2
  • Qdrant + Neo4j (no shared transaction) (composed_qdrant_neo4j), lane e2: neo4j_checkpoint_interval for neo4j_checkpoint is missing, tier e2_500k
  • Qdrant + Neo4j (no shared transaction) (composed_qdrant_neo4j), lane e2: neo4j_pagecache for neo4j_pagecache is missing, tier e2
  • Qdrant + Neo4j (no shared transaction) (composed_qdrant_neo4j), lane e2: neo4j_pagecache for neo4j_pagecache is missing, tier e2_500k
• Rows with no image digest:
  • ArcadeDB (embedded) (arcadedb_graph_embedded), lane l2: the rows carry no image field
  • Chroma (chroma_dense), lane l3d: the rows carry a tag only
  • DuckDB (duckdb), lane l1tpc: the rows carry no image field
  • DuckDB (duckdb), lane l4: the rows carry a tag only
  • DuckDB VSS (duckdb_vss_dense), lane l3d: the rows carry a tag only
  • DuckPGQ (duckpgq_graph), lane l2: the rows carry no image field
  • LadybugDB (ladybug_graph), lane l2: the rows carry no image field
  • LanceDB (lancedb_dense), lane l3d: the rows carry a tag only
  • MongoDB (mongodb_dense), lane l3d: the rows carry a tag only
  • SQLite (sqlite), lane l1tpc: the rows carry no image field
  • SQLite (sqlite), lane l4: the rows carry a tag only
  • SurrealDB (embedded) (surrealdb_e2), lane e2: the rows carry no image field
  • SurrealDB (embedded) (surrealdb_graph), lane l2: the rows carry no image field
  • SurrealDB (embedded) (surrealdb_lifecycle), lane lifecycle: the rows carry a tag only
  • SurrealDB (embedded) (surrealdb_tpc), lane l1tpc: the rows carry no image field
  • SurrealDB (embedded) (surrealdb_ts), lane l4: the rows carry a tag only
  • sqlite-vec (sqlite_vec_dense), lane l3d: the rows carry a tag only
  • sqlite-vec (sqlite_vec_dense_int8), lane l3d: the rows carry a tag only
• Tiers whose rows mix versions:
  • ArangoDB (arangodb_graph), lane l2: sf1, sf10
  • FalkorDB (falkordb_graph), lane l2: sf1, sf10
  • LadybugDB (ladybug_graph), lane l2: sf1, sf10

How to tell us a setting is wrong

If a setting we use for your engine differs from what you recommend for this task, open an issue (see Contribute) with the engine, the version, the setting, and a link to your documentation. A corrected configuration is re-run on the same machine and published beside the earlier result, with the reason for the change.