AiSAQ is an all-in-storage approximate nearest neighbor search system that uses product quantization to enable DRAM-free vector similarity search, serving as a specialized vector search/indexing approach for large-scale information retrieval.
title: AiSAQ slug: aisaq category: Research Papers & Surveys tags:
AiSAQ (All-in-Storage ANNS with Product Quantization) is a research method for approximate nearest neighbor search (ANNS) that places compressed vectors entirely on SSD, enabling DRAM-free (or near-DRAM-free) vector similarity search at billion-scale. It builds on DiskANN, modifying how product-quantized vectors are stored and accessed to drastically cut RAM usage while maintaining high recall and practical latency.
All-in-storage PQ design
Extremely low DRAM footprint
Based on DiskANN
Product Quantization (PQ) for compression
Maintains recall–latency balance
Fast index switching
Loading more......
Suitable for RAG (Retrieval-Augmented Generation)
Scalability and multi-server deployment
Use with vector database systems