PBench is a benchmarking and performance testing toolkit for Presto and Prestissimo query engines.
PBench addresses the need to measure and validate query performance across Presto and Prestissimo clusters. It provides configurable concurrent query workloads that capture results and verify correctness, allowing teams to establish performance baselines and detect regressions. Beyond basic benchmarking, the tool supports workload replay by recording and replaying production query traffic, real-time query forwarding to monitor source clusters and route queries to targets, and schema reproduction through table metadata export. It can generate standard benchmark datasets like TPC-DS and TPC-H, compare query results across runs, and load query metadata into databases for offline analysis.
Teams should adopt PBench when they need comprehensive performance validation for Presto or Prestissimo deployments, particularly those managing production workloads where understanding query behavior across environments matters. It suits organizations that want to move beyond simple query timing to capture full query context, verify result correctness, and track performance changes over time. The tool originated as a replacement for Benchto, incorporating lessons from that project while expanding into operational monitoring and workload analysis capabilities.
The project maintains active development with regular updates to its command-line interface and core functionality. Documentation is extensive, covering installation, configuration through stage JSON files, command usage, and output interpretation. The codebase is written in Go and includes development guidelines for contributors. The project provides both quick-start paths for new users and detailed reference documentation for advanced configuration and analysis workflows.
