G-Store: A Storage Manager for Graph Data

Graph data is ubiquitous: Social networks, Semantic Web, pointer analysis in software engineering, and biological and chemical networks all rely on a graph representation of data. This paper makes the case for a native storage layer for graph data, rather than relying on relational or columnar stores. We propose a lightweight storage manager for graph data called G-Store. It exploits the structure of the graph for placement of data in pages that is optimized for a wide range of access patterns found in graph queries. Our placement approach partitions the data into pages using a multilevel partitioning algorithm and arranges the pages on disk to minimize the distance on disk between adjacent vertices. Initial experiments show that G-Store can outperform existing graph database solutions by orders of magnitude. We believe that these results justify a promising avenue of research into storage-aware graph databases. We discuss some of these research directions.