Publication

SPFresh: Incremental In-Place Update for Billion-Scale Vector Search

High-dimensional vector data approximate nearest neighbor search (ANNS) is widely used in information retrieval, question answering, and recommendation systems. SPFresh introduces LIRE, a lightweight incremental rebalancing protocol, enabling in-place vector updates that maintain search quality without costly full index rebuilds.

SOSP 2023 / October 2023

vector searchANNSin-place updatebillion-scale

PDF

Authors

Yuming Xu, Hengyu Liang, Jin Li, Shuotao Xu, Qi Chen, Qianxi Zhang, Cheng Li, Ziyue Yang, Fan Yang, Yuqing Yang, Peng Cheng, Mao Yang

Abstract

SPFresh introduces LIRE, a lightweight incremental rebalancing protocol for billion-scale vector search, enabling efficient in-place updates that maintain high search quality without full index rebuilds.