Benchmarking your commits: The Postgres Performance Farm
summary
A benchmarking pipeline that automatically clones and builds Postgres, runs pgbench and TPC-H, uploads results to a web server and links each result to the exact commit and branch. A Python client builds Postgres from scratch, runs configurable workloads, collects hardware, OS, compiler and Postgres settings (about 50 tunables), and stores statement latencies, pg_stat_statements and JSON exports. It started in Google Summer of Code, runs on dedicated GSoC servers now, and is usable but needs more runners, bugs and deployment to the main Postgres host.
AI-written summary · about
chapters (8)
- 0:00Introduction and background
- 2:25Project origin and GSoC
- 7:59Defining the performance farm
- 13:12Architecture and workloads
- 18:57Call for contributions
- 23:32Analyzing performance data
- 33:41Future goals and summary
- 37:24Q&A session
related talks
36mAutomated Performance Testing of Postgres
50mPostgreSQL hacker tips
29mPerformance Archaeology - 20 years of improvements
31mDemanding the impossible rigorous database benchm
26mReproducible Postgres
50mPerformance Archaeology
45mTuning Hardware and the Operating System: Benchmarking on PostgreSQL
36mBest Practices in PostgreSQL Tuning: Navigating Key Performance Bottlenecks in the Cloud