Link Search Menu Expand Document Documentation Menu
Documentation

You're viewing version 3.8 of the OpenSearch documentation. This version is no longer maintained. For the latest version, see the current documentation. For information about OpenSearch version maintenance, see Release Schedule and Maintenance Policy.

Optimizing vector storage

Vector search operations can be resource intensive, especially when dealing with large-scale vector datasets. OpenSearch provides several optimization techniques for reducing memory usage.

Vector quantization

Reduce vector storage space by quantizing vectors

Disk-based vector search

Uses binary quantization to reduce the operational costs of vector workloads

Disk-friendly quantization using the opensearch-jvector plugin

The jvector engine, provided by the opensearch-jvector plugin, implements DiskANN-style indexing: it stores vectors on disk rather than in memory and builds indexes directly from quantized vectors. This reduces memory use without requiring separate quantization configuration.

Compared with the built-in engines, the jvector engine provides the following storage advantages:

  • It builds indexes from quantized vectors, reducing the memory required during indexing.
  • It refines quantization codebooks incrementally during merges, with no full rebuilds.
350 characters left

Have a question? .

Want to contribute? or .