GoVector: An I/O-Efficient Caching Strategy for High-Dimensional Vector Nearest Neighbor Search

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhou, Yijie, Lin, Shengyuan, Gong, Shufeng, Yu, Song, Fan, Shuhao, Zhang, Yanfeng, Yu, Ge
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915455726256128
author Zhou, Yijie
Lin, Shengyuan
Gong, Shufeng
Yu, Song
Fan, Shuhao
Zhang, Yanfeng
Yu, Ge
author_facet Zhou, Yijie
Lin, Shengyuan
Gong, Shufeng
Yu, Song
Fan, Shuhao
Zhang, Yanfeng
Yu, Ge
contents Graph-based high-dimensional vector indices have become a mainstream solution for large-scale approximate nearest neighbor search (ANNS). However, their substantial memory footprint often requires storage on secondary devices, where frequent on-demand loading of graph and vector data leads to I/O becoming the dominant bottleneck, accounting for over 90\% of query latency. Existing static caching strategies mitigate this issue only in the initial navigation phase by preloading entry points and multi-hop neighbors, but they fail in the second phase where query-dependent nodes must be dynamically accessed to achieve high recall. We propose GoVector, an I/O-efficient caching strategy tailored for disk-based graph indices. GoVector combines (1) a static cache that stores entry points and frequently accessed neighbors, and (2) a dynamic cache that adaptively captures nodes with high spatial locality during the second search phase. To further align storage layout with similarity-driven search patterns, GoVector reorders nodes on disk so that similar vectors are colocated on the same or adjacent pages, thereby improving locality and reducing I/O overhead. Extensive experiments on multiple public datasets show that GoVector achieves substantial performance improvements. At 90% recall, it reduces I/O operations by 46% on average, increases query throughput by 1.73x, and lowers query latency by 42% compared to state-of-the-art disk-based graph indexing systems.
format Preprint
id arxiv_https___arxiv_org_abs_2508_15694
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GoVector: An I/O-Efficient Caching Strategy for High-Dimensional Vector Nearest Neighbor Search
Zhou, Yijie
Lin, Shengyuan
Gong, Shufeng
Yu, Song
Fan, Shuhao
Zhang, Yanfeng
Yu, Ge
Databases
Graph-based high-dimensional vector indices have become a mainstream solution for large-scale approximate nearest neighbor search (ANNS). However, their substantial memory footprint often requires storage on secondary devices, where frequent on-demand loading of graph and vector data leads to I/O becoming the dominant bottleneck, accounting for over 90\% of query latency. Existing static caching strategies mitigate this issue only in the initial navigation phase by preloading entry points and multi-hop neighbors, but they fail in the second phase where query-dependent nodes must be dynamically accessed to achieve high recall. We propose GoVector, an I/O-efficient caching strategy tailored for disk-based graph indices. GoVector combines (1) a static cache that stores entry points and frequently accessed neighbors, and (2) a dynamic cache that adaptively captures nodes with high spatial locality during the second search phase. To further align storage layout with similarity-driven search patterns, GoVector reorders nodes on disk so that similar vectors are colocated on the same or adjacent pages, thereby improving locality and reducing I/O overhead. Extensive experiments on multiple public datasets show that GoVector achieves substantial performance improvements. At 90% recall, it reduces I/O operations by 46% on average, increases query throughput by 1.73x, and lowers query latency by 42% compared to state-of-the-art disk-based graph indexing systems.
title GoVector: An I/O-Efficient Caching Strategy for High-Dimensional Vector Nearest Neighbor Search
topic Databases
url https://arxiv.org/abs/2508.15694