Skip to content
Steve Burnett edited this page Feb 25, 2026 · 29 revisions

PBench

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.

Getting Started

Writing Benchmarks

Included Benchmarks

  • Included Benchmarks — TPC-H, TPC-DS, ClickBench, IMDB/JOB queries and configs shipped with PBench

Running Benchmarks

Errors

All Commands

Contributing

  • Development — building, testing, and contributing to PBench

Clone this wiki locally