Search latency on object storage
Full-text, vector and hybrid latency from the engine's own benchmark harness, measured in CI on 1M-document tables in Azure Blob with a warm cache.
Results
| query | shape | result |
|---|---|---|
| full-text (BM25) | single rare term, top-10, warm + fetch | 166 µs p50 |
| vector | top-10 cosine, 1024 dimensions, post-drain | 663 µs p50 · 0.992 recall@10 |
| hybrid | hybrid_search through SQL, 10 rows | 3.47 ms p50 |
| ingest | full-text / vector / SQL tables | 62.1 / 7.8 / 17.0 K docs/s |
- storage
- Azure Blob, 1M-document tables, warm cache
- hardware
- Intel Xeon 6973P-C, 4 cores pinned, 31 GiB
- engine
- commit bd7caa3, measured in CI
- harness
- cargo bench in the engine repository
What the numbers cover
- The table lives in object storage. Data and indexes are in Azure Blob, and the engine reads byte ranges from it, with hot ranges cached on local disk.
- Warm cache. These are repeat queries after the first touch. The first query to touch a file pays object-store round trips, covered in tradeoffs.
- One small machine. Four pinned cores, so the numbers are per-query latency, not a cluster's throughput.
- Every cell has a run. The commit, the CI run and the raw logs sit next to each number in benches/README.md.
Reproduce it
The harness ships with the engine. Every cell and the flags that select it are listed in benches/README.md.
git clone https://github.com/infino-ai/infino && cd infino cargo bench
Common questions
What does a query cost the first time it touches object storage?
The first query to touch a file pays object-store round trips, hundreds of milliseconds, before its bytes are cached. Warm and repeated queries do not pay it. The figures on this page are warm, and the tradeoffs page in the docs states the cold envelope.
How is this different from ClickBench?
ClickBench measures SQL analytics over one 100M-row dataset on local disk, run by ClickBench. This page measures search: full-text, vector and hybrid queries against tables that live in object storage, run by the engine's own harness in CI.
Keep the data in your bucket. Search it in microseconds.
Free tier · No credit card to start
or run it yourself: pip install infino · Apache-2.0 · quickstart →