// BENCHMARKS

Real numbers. Real hardware.

Every number on this page was measured on real hardware with --release builds. Random-vector benchmarks are not reported. Commands to reproduce every result are shown inline.

Hardware: AWS c6id.4xlarge·Intel Xeon Platinum 8375C, 16 vCPU, 32 GiB RAM, 884 GB NVMe·Ubuntu 24.04, io_uring backend

HNSW Vector Search -- SIFT-1M

Dataset: SIFT-1M -- 1,000,000 x 128-dim float32 vectors, 10,000 queries, pre-computed ground truth. Source: ftp://ftp.irisa.fr/local/texmex/corpus/sift.tar.gz
SHA256: 92f1270c5e3a0cb46b89983e72b0511e4df065c31a9fa0276d8c9b1fca5bc81a

Build: M=16, ef_construction=200 -- 66.2s (15,114 vec/sec)
ef_searchrecall@10mean latencyp99 latency
100.76257.6 µs101 µs
500.959158.1 µs228 µs
1000.983266.5 µs364 µs
2000.990459.4 µs616 µs
# Reproduce
cargo run --release -p galaxdb-benchmarks -- \
--sift-dir /path/to/sift \
--ef-sweep 10,50,100,200 \
--output bench-results/sift_bench.json

Storage Engine

Write throughput

WorkloadThroughputDurability
OLTP -- 16 threads, 1M rows, 60s258,555 TPSRelaxed (group commit)
Embedded INSERT -- batched 100/stmt20,267 rows/secRelaxed
Wire INSERT -- 4 clients454 rows/secStrict (fsync/commit)

Read latency (warm cache)

Point read p503 µs
Point read p9947 µs

Read latency (cold cache, 50M rows x 600B)

Missing keys0 / 100,000
Read p50147 µs
Read p99308 µs

OLAP scan throughput

Scan throughput4.49 GB/s
Zone-map skip rate80%

Mixed OLTP + OLAP (concurrent, 60s)

OLTP p99 during concurrent scan191 µs -- no degradation
HotSet evictions from scan storm0 -- ScanBuffer isolation works

Encryption

Measured with cargo bench -p galaxdb-crypto on the same hardware.

OperationLatencyThroughput
AEGIS-256 decrypt 1 MB151 µs6.63 GB/s
AEGIS-256 encrypt 64 KB9.75 µs6.56 GB/s
AES-256-GCM decrypt 1 MB701 µs1.43 GB/s
XXH3-64 checksum 1 MB--34.1 GB/s
ART lookup (1M keys)168 ns/op--

Crash Safety

All 7 chaos scenarios pass in under 30s total.

ScenarioResultTime
Kill mid-flush -- WAL replay1,000 rows recovered, zero loss8.79s
Kill mid-compaction -- old blocks intact4,000 keys readable0.02s
Corrupt WAL record -- replay stops at corruptionPartial recovery, no corrupt data returned1.81s
Disk full -- clean checkpoint, writes blockedReserve file freed, reads continue0.01s
Kill sidecar -- backlog preserved, no data loss50 requests queued, drained on recovery0.00s
100 concurrent writers100K writes, 0 duplicates, 0 missing0.13s
OLAP scan during OLTP0 HotSet evictions, OLTP p99 unaffected0.15s
Total7 passed / 0 failed10.91s
# Reproduce
cargo run --release -p galaxdb-chaos-tests

Test Suite

740
Rust unit tests
0 failed
7
Chaos scenarios
0 failed
10.91s
Total chaos time
all recovery < 30s
# Reproduce
cargo test --workspace --exclude galaxdb-python --lib

Methodology

  • --All numbers measured on AWS c6id.4xlarge (Intel Xeon Platinum 8375C @ 2.90GHz, 16 vCPU, 32 GiB RAM, 884 GB NVMe), Ubuntu 24.04.
  • --All builds use cargo build --release. Debug builds are never published.
  • --Random-vector HNSW benchmarks are not reported. SIFT-1M only.
  • --Raw benchmark JSON: sift_bench.json
  • --Full reproduction guide: BENCHMARKS.md