A bi-directional shortest path calculation speed up technique for RDBMS

Abstract

In Relational Database Management Systems (RDBMS) the representation and storage of graph datasets is achieved by using two data tables. One representing the nodes’ data and another for the edges’ data of the graph. Using this model every search process in a big data volume requires multiple joins of large data tables, which significantly reduces their performance. The increasing use of graph datasets in various computer science fields such as (big data, social and computer networks, networks, Semantic Web, knowledge graphs, biological networks, and road networks) creates the need to research methods that will improve the performance of RDBMS when managing this type of data. In this paper, unlike most of the solutions proposed in the literature to date, a purely relational approach is presented using only the native RDBMS query analyzer and not either aggregate functions not implemented in all RDBMSs or an additional plugin layer that analyzes and processes the graph queries in order to translate them into the corresponding SQL versions. With the proposed solution we proved that by designing a bi-directional BFS algorithm that passes through properly selected intermediate pivot points, it is possible to query graph datasets significantly larger than those a complex SQL query can handle, with no major overhead.