{"id":5462,"date":"2026-06-08T10:52:06","date_gmt":"2026-06-08T17:52:06","guid":{"rendered":"https:\/\/www.couchbase.com\/blog\/?p=5462"},"modified":"2026-06-08T10:52:07","modified_gmt":"2026-06-08T17:52:07","slug":"diskann","status":"publish","type":"post","link":"https:\/\/www.couchbase.com\/blog\/pt\/diskann\/","title":{"rendered":"What Is DiskANN? Billion-Scale Vector Search Explained"},"content":{"rendered":"

Retrieval-augmented generation (RAG), semantic search, and AI agents all depend on one thing: the ability to quickly find the most relevant vectors in a large dataset. As embedding datasets grow from millions to billions of records, purely in-memory vector indexes become financially unsustainable. DiskANN solves this problem by storing vector indexes on SSD rather than RAM, enabling web-scale search on commodity hardware.<\/p>\n\n\n\n

What is DiskANN?<\/h2>\n\n\n\n

DiskANN <\/strong>is a graph-based vector search algorithm developed by Microsoft Research that was first published at NeurIPS 2019. It performs approximate nearest neighbor (ANN) search over billion-scale vector datasets, using SSD as its primary storage medium and keeping only a compressed representation of the index in RAM.<\/p>\n\n\n\n

Before DiskANN, algorithms like HNSW and FAISS required the entire vector index to reside in DRAM. At a billion vectors, that demands hundreds of gigabytes of RAM and an infrastructure that\u2019s expensive to provision and scale. DiskANN breaks this constraint by shifting the bulk of the index to disk while maintaining recall and latency characteristics competitive with in-memory approaches. It\u2019s built on top of the Vamana algorithm, a novel graph construction method that produces a single-layer directed graph well-suited for efficient disk-based traversal.<\/p>\n\n\n\n

Key benchmarks from the original <\/strong>NeurIPS 2019 paper<\/strong><\/a>:<\/strong><\/p>\n\n\n\n