Graph Reordering for Cache-Efficient Near Neighbor Search

TL;DR

This paper investigates how graph reordering can improve cache efficiency in near neighbor search algorithms, demonstrating up to 40% faster query times with negligible reordering overhead.

Contribution

It introduces the application of graph reordering algorithms to near neighbor graphs, significantly enhancing cache performance and query speed.

Findings

Reordering improves query time by up to 40%.

Reordering overhead is negligible compared to index construction.

Reordering enhances cache efficiency in graph-based search algorithms.

Abstract

Graph search is one of the most successful algorithmic trends in near neighbor search. Several of the most popular and empirically successful algorithms are, at their core, a simple walk along a pruned near neighbor graph. Such algorithms consistently perform at the top of industrial speed benchmarks for applications such as embedding search. However, graph traversal applications often suffer from poor memory access patterns, and near neighbor search is no exception to this rule. Our measurements show that popular search indices such as the hierarchical navigable small-world graph (HNSW) can have poor cache miss performance. To address this problem, we apply graph reordering algorithms to near neighbor graphs. Graph reordering is a memory layout optimization that groups commonly-accessed nodes together in memory. We present exhaustive experiments applying several reordering algorithms to a leading graph-based near neighbor method based on the HNSW index. We find that reordering improves the query time by up to 40%, and we demonstrate that the time needed to reorder the graph is negligible compared to the time required to construct the index.