GalaxyView: Property Graph Transformation for Materialized View Generation

In the practical use of graph databases, storing graphs separately enhances maintainability, while integrating them into a unified graph facilitates advanced analytics. To address these dual needs, we present a GQL-compatible framework for creating graph views across multiple property graphs. Leveraging insights from commercial graph database environments and user-driven requirements, we identify two key graph transformations - merging and expansion - that significantly boost query efficiency. By incorporating these transformations into view generation, we effectively minimize redundant queries. Furthermore, we streamline view creation through implicit edge creation and automated property merging. To ensure views meet user expectations before full generation, we introduce a preview feature that uses sampling to effectively represent the graph structure. We assess the performance of GalaxyView across diverse workloads, with comprehensive experiments on real-world graphs demonstrating significant improvements in query performance and operational efficiency.