PBench is a toolkit for Presto and Trino performance testing and operations. Originally a replacement for Benchto, PBench has grown to include:

• Benchmarking — run configurable, concurrent query workloads with results capturing and correctness verification
• Workload replay — replay recorded production query traffic against a cluster
• Query forwarding — monitor a source cluster and forward queries to one or more target clusters in real time
• Table schema saving — export table metadata and statistics for reproducing schemas across environments
• Result comparison — diff query output directories to detect regressions
• DDL generation — generate CREATE/INSERT scripts for benchmark datasets (TPC-DS, TPC-H, etc.)
• Query JSON analysis — load query info JSON files into databases for offline analysis

For a detailed explanation of PBench's design and architecture, see Comprehensive Performance Benchmarking, Monitoring, and Reporting Infrastructure for Presto and Prestissimo.

• QuickStart — install and run your first benchmark in minutes
• Installing PBench — platform-specific download and setup

• Configuring PBench — create stage JSON files to define queries, sessions, and execution order
  • Parameters — complete reference for all stage JSON parameters

• The Run Command — flags and options for pbench run
  • Understanding PBench Output — output directory structure and result files
  • Comparing Results — use pbench cmp to diff query output directories

• Command Reference — online help for every pbench subcommand (run, cmp, forward, replay, save, etc.)
• Generating Benchmark Configurations — use pbench genconfig to generate cluster configs from templates

• Development — building, testing, and contributing to PBench