Gorgeous: Revisiting the Data Layout for Disk-Resident High-Dimensional Vector Search

2025-08-21

AI Summary

To address low cache efficiency and poor disk access locality in TB-scale high-dimensional vector similarity search on SSDs—caused by co-locating graph indices and vector data—this paper is the first to explicitly distinguish the distinct memory access patterns: proximity graphs (high-frequency, random access) versus raw vectors (low-frequency, bulk access). We propose a graph-centric data layout and caching optimization: (1) a compact adjacency-list memory caching mechanism for graph structures, and (2) a redesigned on-disk block format to enhance I/O locality. Experiments under disk-resident settings demonstrate that our approach achieves over 60% higher average query throughput and over 35% lower latency compared to state-of-the-art systems, significantly alleviating the I/O bottleneck in large-scale vector retrieval.

Technology Category

• Search and Optimization: Distributed Search
• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems

Application Category

• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Security and Privacy: Large-scale security measurements

Abstract

Similarity-based vector search underpins many important applications, but a key challenge is processing massive vector datasets (e.g., in TBs). To reduce costs, some systems utilize SSDs as the primary data storage. They employ a proximity graph, which connects similar vectors to form a graph and is the state-of-the-art index for vector search. However, these systems are hindered by sub-optimal data layouts that fail to effectively utilize valuable memory space to reduce disk access and suffer from poor locality for accessing disk-resident data. Through extensive profiling and analysis, we found that the structure of the proximity graph index is accessed more frequently than the vectors themselves, yet existing systems do not distinguish between the two. To address this problem, we design the Gorgeous system with the principle of prioritizing graph structure over vectors. Specifically, Gorgeous features a memory cache that keeps the adjacency lists of graph nodes to improve cache hits and a disk block format that explicitly stores neighbors' adjacency lists along with a vector to enhance data locality. Experimental results show that Gorgeous consistently outperforms two state-of-the-art disk-based systems for vector search, boosting average query throughput by over 60% and reducing query latency by over 35%.

Problem

• Optimizing data layout for disk-resident high-dimensional vector search
• Reducing disk access through improved memory utilization strategies
• Enhancing data locality for proximity graph index structures

Innovation

• Prioritizes graph structure over vectors
• Memory cache for adjacency lists
• Disk block format enhances data locality