In-Place Updates of a Graph Index for Streaming Approximate Nearest Neighbor Search

TL;DR

This paper introduces IP-DiskANN, an innovative graph index update algorithm that processes insertions and deletions in-place, maintaining stable recall and outperforming batch methods in streaming approximate nearest neighbor search scenarios.

Contribution

IP-DiskANN is the first algorithm to enable in-place updates for graph indices in streaming ANNS, eliminating the need for batch consolidation while maintaining high recall stability.

Findings

IP-DiskANN maintains stable recall over long update sequences.

It achieves higher query throughput and update speed than batch consolidation methods.

Outperforms HNSW in various benchmarks.

Abstract

Indices for approximate nearest neighbor search (ANNS) are a basic component for information retrieval and widely used in database, search, recommendation and RAG systems. In these scenarios, documents or other objects are inserted into and deleted from the working set at a high rate, requiring a stream of updates to the vector index. Algorithms based on proximity graph indices are the most efficient indices for ANNS, winning many benchmark competitions. However, it is challenging to update such graph index at a high rate, while supporting stable recall after many updates. Since the graph is singly-linked, deletions are hard because there is no fast way to find in-neighbors of a deleted vertex. Therefore, to update the graph, state-of-the-art algorithms such as FreshDiskANN accumulate deletions in a batch and periodically consolidate, removing edges to deleted vertices and modifying the graph to ensure recall stability. In this paper, we present IP-DiskANN (InPlaceUpdate-DiskANN), the first algorithm to avoid batch consolidation by efficiently processing each insertion and deletion in-place. Our experiments using standard benchmarks show that IP-DiskANN has stable recall over various lengthy update patterns in both high-recall and low-recall regimes. Further, its query throughput and update speed are better than using the batch consolidation algorithm and HNSW.