Graph Database Performance Benchmark
PropGBench is the first benchmark of its kind to define comprehensive query categories for property graphs and evaluate the performance of different graph database systems across them. Its query suite covers most real-world graph database query scenarios, while its metrics, including runtime, peak memory usage, and I/O usage, provide a thorough and comprehensive view of system performance.
Workload Categories
PropGBench groups its benchmark queries into five workload categories, from local traversals to global graph algorithms.
Neighborhood
Measures local traversals such as k-hop, exact-kth-hop, and filtered neighbor lookups.
Path
Covers path and reachability workloads, such as reachability checks, shortest paths, and bounded weighted path enumerations.
Pattern
Exercises structural pattern matching with graph motifs and multi-step joins, such as triangles, star patterns, and two-hop joins.
Aggregation
Summarizes matched graph patterns with grouped counts, degree statistics, and top-K results.
Analytics
Runs graph analytic algorithms, such as degree centrality, PageRank, connected components, and community detection.
Metrics
To introduce the metrics we used to evaluate the performance of each graph engine.
Sponsors
