On isomorphic matching of large disk-resident graphs using an XQuery engine

There exists an increasing interest in using graphs to model data, and managing them is a challenging research field. One of the major hurdles in large graph management and processing is our ability to store graphs on disk, and develop techniques that can process the data in their native representation on the disk. Currently, many powerful processing techniques only ensure efficient processing while the graphs reside fully in volatile memory, which limits their applications. In this paper, we present a disk representation of unit graphs, called graphlets, that is amenable to leveraging both XML and relational storage structures, and associated query engines such as XQuery and SQL3. Specifically, we focus on XML and XQuery to implement a graph decomposition-based isomorphic subgraph matching technique, called NetQL, that exploits the graphlet representation. Furthermore, we present a new covering concept, called the minimum hub cover, that allows node-at-a-time processing of arbitrarily large graphs and opens up new opportunities for cost-based graph query optimization. Finally, we discuss some early results to show that such optimizations are feasible and promising by comparing our strategy with GraphQL.