Achieving Low-Latency Graph-Based Vector Search via Aligning Best-First Search Algorithm with SSD

Learn about PipeANN, an innovative on-disk graph-based approximate nearest neighbor search (ANNS) system that dramatically reduces the latency gap between disk-based and in-memory vector search solutions in this 12-minute conference presentation from OSDI '25. Discover how researchers from Tsinghua University aligned the best-first search algorithm with SSD characteristics to avoid strict compute-I/O ordering across search steps, enabling significant performance improvements. Explore the experimental results demonstrating that PipeANN achieves 1.14×--2.02× search latency compared to in-memory Vamana and only 35.0% of the latency of on-disk DiskANN in billion-scale datasets, all while maintaining search accuracy. Gain insights into the technical innovations that make high-performance vector search feasible on storage devices, addressing critical scalability challenges in large-scale machine learning and information retrieval systems.