U1MIndexer
An indexer that leverages hnswlib for vector search and DocumentArrayMemmap for storing full Documents. U1M means under one million Documents. It is a perfect indexer if you are working with indexers less than 1M and aims for query-time speed. When working with one million Documents, you can expect single query to be complete at 2ms and 600~800 QPS on batch queries. When working with more than 1M Documents, one has to be careful with its memory consumption. The indexing speed of U1MIndexer is much slower comparing to U100KIndexer.
U1MIndexer returns approximate nearest neighbours.
The code is based on HNSWSearcher
Pros & cons
Pros
- Extremely fast query speed, irrelevant to the number of stored Documents.
- Always return full Documents.
Cons
- Indexing is relatively slow.
- Extra dependencies:
hnswlibandbidict. - Extra parameters to tune HNSW performance.
Performance
One can run benchmark.py to get a quick performance overview.
| Stored data | Indexing time | Query size=1 | Query size=8 | Query size=64 |
|---|---|---|---|---|
| 10000 | 2.903 | 0.002 | 0.013 | 0.092 |
| 100000 | 65.330 | 0.002 | 0.014 | 0.104 |
| 500000 | 463.509 | 0.003 | 0.019 | 0.132 |
| 1000000 | 988.535 | 0.003 | 0.017 | 0.127 |