Applying Incremental Graph Transformation to Existing Models in Relational Databases

Abstract

As industrial practice demands larger and larger system models, the efficient execution of graph transformation remains an important challenge. Additionally, for real-world applications, compatibility and integration with already well-established technologies is highly desirable. Therefore, relational databases have been investigated before as off-the-shelf environments for graph transformation, since they are already widely used for storing, processing and querying large graphs. The graph pattern matching phase of graph transformation typically dominates in cost due to its combinatorial complexity. Therefore significant attempts have been made to improve this process; incremental pattern matching is an approach that has been shown to exert favorable performance characteristics in many practical use cases. To this day, however, no solutions are available for applying incremental techniques side by side with already deployed systems built over relational databases. In the current paper, we propose an approach that translates graph patterns and transformation rules into event-driven (trigger-based) SQL programs that seamlessly integrate with existing relational databases to perform incremental pattern matching. Additionally, we provide experimental evaluation of the performance of our approach.