Pinned post
If faster vector search, reduced storage requirements, and more powerful observability are on your list, OpenSearch 3.9 is worth a look.
Key updates include:
• Neural sparse ANN search with 39% higher throughput, 3.3x faster index builds, and an 8x smaller JVM heap in testing
• Native half_float vectors that can reduce vector memory and storage by roughly 50%
• New 2-bit and 4-bit scalar quantization options for tuning compression vs. recall
• A unified view for alerts, anomalies, and forecasts
• Adaptive per-action concurrency limits for protecting clusters under load
• Index-level search pruning that reduced tail latency by more than 80% and can_match requests by nearly 98% in local benchmarks
Read the blog for the full breakdown of what’s new in OpenSearch 3.9: opensearch.org/blog/get-to-kn…
Key updates include:
• Neural sparse ANN search with 39% higher throughput, 3.3x faster index builds, and an 8x smaller JVM heap in testing
• Native half_float vectors that can reduce vector memory and storage by roughly 50%
• New 2-bit and 4-bit scalar quantization options for tuning compression vs. recall
• A unified view for alerts, anomalies, and forecasts
• Adaptive per-action concurrency limits for protecting clusters under load
• Index-level search pruning that reduced tail latency by more than 80% and can_match requests by nearly 98% in local benchmarks
Read the blog for the full breakdown of what’s new in OpenSearch 3.9: opensearch.org/blog/get-to-kn…
