Improved Multitenancy Support in Vector Search: Introducing Flat Indexes

The future of AI is becoming increasingly personalized, with users desiring tailored experiences that rely on specific memory capabilities for individual or business needs, such as AI assistants or legal tools. This trend necessitates multitenancy, where a single software instance serves multiple isolated customers, but this has historically posed scaling and performance challenges, particularly with vector searches. To address these issues, MongoDB Atlas introduces Flat Indexes for Vector Search, offering enhanced support for multitenant workloads by improving performance, recall, and resource efficiency compared to the traditional Hierarchical Navigable Small World (HNSW) algorithm. Flat Indexes optimize for highly selective queries by performing exhaustive scans of data specific to a tenant, thereby eliminating the noisy neighbor problem and providing predictable performance. This new indexing method is ideal for applications with many tenants each having a small number of vectors, whereas HNSW remains suitable for larger shared datasets. Flat Indexes, now available in Public Preview, represent a step towards making MongoDB a leading platform for multitenant AI applications, with future enhancements planned.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.