...

Elastic Introduces New Vector Storage Format DiskBBQ for More Efficient Vector Search

Vector Storage Format

Elastic announced its new vector storage format called “DiskBBQ”, designed to optimize vector search at scale. In effect, Elastic is advancing vector search infrastructure by offering a format that uses far less memory and delivers predictable performance.

Traditionally, many vector databases rely on the Hierarchical Navigable Small Worlds (HNSW) algorithm for similarity search. However, HNSW requires all vectors in memory, which becomes costly at a large scale. Elastic’s DiskBBQ changes that by compressing vectors and partitioning them into clusters so disk reads become selective rather than loading everything into RAM. 

In benchmark testing, DiskBBQ sustained query latencies around 15 milliseconds while using as little as 100 MB of total memory – a scenario where HNSW could not run. This disk-friendly, approximate nearest-neighbor (ANN) algorithm removes memory as a limiting factor and allows datasets to scale based only on CPU and disk. 

Elastic’s general manager of Platform, Ajay Nair, commented, “As AI applications scale, traditional vector storage formats force them to choose between slow indexing or significant infrastructure costs required to overcome memory limitations.” The company says DiskBBQ is available now in Elasticsearch 9.2 for technical preview in Elasticsearch Serverless. 

By introducing DiskBBQ, Elastic addresses key industry challenges – large-scale vector search, infrastructure cost, stability under load, and resource optimization. As AI workloads increasingly demand performant vector search, Elastic’s move could shift how enterprises design their search stacks.

Going forward, organizations using vector embeddings for search, recommendation, or AI applications can benefit from DiskBBQ. It allows them to store and query massive vector indexes with significantly lower RAM usage. With the format’s ability to scale without massive memory investment, it may become a new standard in vector‐search engineering.

Explore IT Tech News for the latest advancements in Information Technology & insightful updates from industry experts!

News Source: Businesswire.com

Share with friends

Latest News