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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Yuming, Liang, Hengyu, Li, Jin, Xu, Shuotao, Chen, Qi, Zhang, Qianxi, Li, Cheng, Yang, Ziyue, Yang, Fan, Yang, Yuqing, Cheng, Peng, Yang, Mao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910656790265856
author Xu, Yuming
Liang, Hengyu
Li, Jin
Xu, Shuotao
Chen, Qi
Zhang, Qianxi
Li, Cheng
Yang, Ziyue
Yang, Fan
Yang, Yuqing
Cheng, Peng
Yang, Mao
author_facet Xu, Yuming
Liang, Hengyu
Li, Jin
Xu, Shuotao
Chen, Qi
Zhang, Qianxi
Li, Cheng
Yang, Ziyue
Yang, Fan
Yang, Yuqing
Cheng, Peng
Yang, Mao
contents Approximate Nearest Neighbor Search (ANNS) is now widely used in various applications, ranging from information retrieval, question answering, and recommendation, to search for similar high-dimensional vectors. As the amount of vector data grows continuously, it becomes important to support updates to vector index, the enabling technique that allows for efficient and accurate ANNS on vectors. Because of the curse of high dimensionality, it is often costly to identify the right neighbors of a single new vector, a necessary process for index update. To amortize update costs, existing systems maintain a secondary index to accumulate updates, which are merged by the main index by global rebuilding the entire index periodically. However, this approach has high fluctuations of search latency and accuracy, not even to mention that it requires substantial resources and is extremely time-consuming for rebuilds. We introduce SPFresh, a system that supports in-place vector updates. At the heart of SPFresh is LIRE, a lightweight incremental rebalancing protocol to split vector partitions and reassign vectors in the nearby partitions to adapt to data distribution shift. LIRE achieves low-overhead vector updates by only reassigning vectors at the boundary between partitions, where in a high-quality vector index the amount of such vectors are deemed small. With LIRE, SPFresh provides superior query latency and accuracy to solutions based on global rebuild, with only 1% of DRAM and less than 10% cores needed at the peak compared to the state-of-the-art, in a billion scale vector index with 1% of daily vector update rate.
format Preprint
id arxiv_https___arxiv_org_abs_2410_14452
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle SPFresh: Incremental In-Place Update for Billion-Scale Vector Search
Xu, Yuming
Liang, Hengyu
Li, Jin
Xu, Shuotao
Chen, Qi
Zhang, Qianxi
Li, Cheng
Yang, Ziyue
Yang, Fan
Yang, Yuqing
Cheng, Peng
Yang, Mao
Information Retrieval
Approximate Nearest Neighbor Search (ANNS) is now widely used in various applications, ranging from information retrieval, question answering, and recommendation, to search for similar high-dimensional vectors. As the amount of vector data grows continuously, it becomes important to support updates to vector index, the enabling technique that allows for efficient and accurate ANNS on vectors. Because of the curse of high dimensionality, it is often costly to identify the right neighbors of a single new vector, a necessary process for index update. To amortize update costs, existing systems maintain a secondary index to accumulate updates, which are merged by the main index by global rebuilding the entire index periodically. However, this approach has high fluctuations of search latency and accuracy, not even to mention that it requires substantial resources and is extremely time-consuming for rebuilds. We introduce SPFresh, a system that supports in-place vector updates. At the heart of SPFresh is LIRE, a lightweight incremental rebalancing protocol to split vector partitions and reassign vectors in the nearby partitions to adapt to data distribution shift. LIRE achieves low-overhead vector updates by only reassigning vectors at the boundary between partitions, where in a high-quality vector index the amount of such vectors are deemed small. With LIRE, SPFresh provides superior query latency and accuracy to solutions based on global rebuild, with only 1% of DRAM and less than 10% cores needed at the peak compared to the state-of-the-art, in a billion scale vector index with 1% of daily vector update rate.
title SPFresh: Incremental In-Place Update for Billion-Scale Vector Search
topic Information Retrieval
url https://arxiv.org/abs/2410.14452