SciDG: Benchmarking Scientific Dynamic Graph Queries

Abstract

Dynamic graphs are increasingly being utilized in domain knowledge modeling and large-scale scientific data management. Managing dynamic graph data requires a graph database system that can handle constantly changing volumes and data versions, while maintaining an acceptable query latency related to versioning. To understand how the design of storage structures affects database performance and assist scientific application developers in finding the optimal storage structure for their dynamic graph application scenarios, we have designed an easy-to-use benchmark framework called SciDG. We also conducted a study on the latencies of five fundamental version-related queries for various scientific application scenarios using SciDG. We evaluated the performance of databases based on three distinct storage principles: Sp-DB, Dp-DB, and Tp-DB. The experimental results indicate that SciDG is a valuable tool for assessing the strengths and weaknesses of different storage structures for dynamic graphs in various scenarios. Additionally, it assists scientists in selecting the most suitable dynamic graph database system for their work.