On Efficiently Equi-Joining Graphs

Despite the growing popularity of techniques related to graph summarization, a general operator for joining graphs on both the vertices and the edges is still missing. Current languages such as Cypher and SPARQL express binary joins through the non-scalable and inefficient composition of multiple traversal and graph creation operations. In this paper, we propose an efficient equi-join algorithm that is able to perform vertex and path joins over a secondary memory indexed graph, also the resulting graph is serialised in secondary memory. The results show that the implementation of the proposed model outperforms solutions based on graphs, such as Neo4J and Virtuoso, and the relational model, such as PostgreSQL. Moreover, we propose two ways how edges can be combined, namely the conjunctive and disjunctive semantics, Preliminary experiments on the graph conjunctive join are also carried out with incremental updates, thus suggesting that our solution outperforms materialized views over PostgreSQL.