Disk-optimized vector engine now available on the Amazon OpenSearch Service

Amazon OpenSearch Service has introduced a new disk-optimized vector engine for k-NN (vector) search with significant cost and performance benefits:

• Enables running modern search applications at one-third of the previous cost
• Optimized for low memory environments through disk mode configuration
• Uses compression techniques like binary quantization
• Maintains search quality through a disk-optimized rescoring mechanism
• Provides low hundreds-of-milliseconds latency

This new feature offers customers a cost-efficient alternative to in-memory vector search, ideal for workloads that do not require single-digit latency. It provides a balance between performance, accuracy, and cost-effectiveness for vector search applications.