HNSW vs. LSH: How Elasticsearch hits 0.99 recall@10 at 15,000 QPS — and what it costs
Elastic Blog - Elasticsearch, Kibana, and ELK Stack

HNSW vs. LSH: How Elasticsearch hits 0.99 recall@10 at 15,000 QPS — and what it costs


Summary

This article explains how Elasticsearch achieves high-performance vector search using HNSW indexing, reaching 15,000 QPS with a 0.99 recall@10 for float32 vectors. It highlights how implementing DiskBBQ quantization can further boost throughput to 55,000 QPS with only a minimal reduction in recall by significantly reducing memory overhead. Ultimately, the text explores the technical trade-offs between search accuracy, speed, and memory usage when managing high-dimensional datasets.
Read the Original Article

This article originally appeared on Elastic Blog - Elasticsearch, Kibana, and ELK Stack.

Read Full Article on Original Site

Popular from Elastic Blog - Elasticsearch, Kibana, and ELK Stack

1
Elastic Stack 9.4.1 released
Elastic Stack 9.4.1 released

adrian brown May 13, 2026 73 views

2
Elastic GenAI Partner Sellers Initiative
Elastic GenAI Partner Sellers Initiative

Sunnie Weber Dec 11, 2025 71 views

3
Elastic Cloud Hosted achieves FedRAMP® High authorization
Elastic Cloud Hosted achieves FedRAMP® High authorization

Chris Townsend Mar 31, 2026 62 views

5
Why AI won’t steal your SOC analyst job
Why AI won’t steal your SOC analyst job

Peter Weller Apr 16, 2026 50 views