We introduce LD-RPQB, a novel benchmark specifically designed for evaluating the performance of Regular Path Queries (RPQs) based on Length distribution for knowledge graphs. RPQs, as a predominant form of navigational queries, allow for the retrieval of vertex pairs connected by paths that match regular expressions. Due to their fundamental position in knowledge graph querying, RPQs have become the focus of extensive research efforts. However, despite their significance, the lack of a comprehensive benchmark for RPQs has hindered realistic performance evaluations. To address this gap, LD-RPQB constructs a synthetic data graph built upon the SP \(^{\varvec{2}}\) Bench framework, with path lengths adjusted to specific distribution patterns, thus ensuring the generated data graph reflects real-world characteristics. In addition, LD-RPQB incorporates 12 query templates derived from both a statistical analysis of real-world corpora and a review of existing research on RPQs. This combination ensures that the benchmark is both scalable and representative of typical query workloads encountered in practical applications. LD-RPQB has been successfully applied to state-of-the-art graph database systems, demonstrating its effectiveness in benchmarking RPQ performance and driving system-level optimizations.
