A Framework for SQL-Based Mining of Large Graphs on Relational Databases

摘要

We design and develop an SQL-based approach for querying and mining large graphs within a relational database management system (RDBMS). We propose a simple lightweight framework to integrate graph applications with the RDBMS through a tightly-coupled network layer, thereby leveraging efficient features of modern databases. Comparisons with straight-up main memory implementations of two kernels -breadth-first search and quasi clique detection - reveal that SQL implementations offer an attractive option in terms of productivity and performance.