This topic describes the types of vector indexes supported by OceanBase Database.
Index types
OceanBase Database supports dense vector indexes, including the HNSW and IVF series. For convenience, these are referred to as dense indexes throughout this document.
| Index type | Description |
|---|---|
| HNSW | The index column supports up to 4096 dimensions. HNSW indexes are in-memory indexes and must be fully loaded into memory. |
| HNSW_SQ | HNSW_SQ indexes offer similar construction speed, search performance, and recall rate as HNSW indexes, but reduce total memory usage to between one-half and one-third of the original. |
| HNSW_BQ | HNSW_BQ indexes have slightly lower recall rates than HNSW indexes but significantly reduce memory usage. The BQ quantization compression algorithm (Rabitq) can compress vectors to 1/32 of their original size. As vector dimensions increase, the memory savings from HNSW_BQ indexes become even more significant. |
| IVF | IVF indexes are implemented on database tables and do not occupy resident memory. |
| IVF_PQ | IVF_PQ indexes are also implemented on database tables and do not occupy resident memory. PQ quantization technology is applied on top of IVF, resulting in slightly lower recall rates but higher performance compared to IVF indexes. In most cases, the PQ quantization compression algorithm can compress vectors to between 1/16 and 1/32 of their original size. |
Considerations and limitations
- Distance algorithms: Dense vector indexes support L2, inner product (IP), and cosine distance algorithms.
- Distance functions: Vector index searches support certain distance functions. For details, see Use SQL functions.
- Filter conditions: Vector searches support filter conditions, which can be scalar conditions or spatial relationships such as ST_Intersects. Multi-value indexes, full-text indexes, and global indexes cannot be used as pre-filters.
- Hybrid search: You can create both vector indexes and full-text indexes on the same table.
- Offline DDL: For information about vector index support for offline DDL operations, see Offline DDL.
- Columnstore indexes: As of V4.3.5 BP3, columnstore vector indexes are not supported.