Performance Archaeology
summary
Benchmarks reconstruct PostgreSQL performance from 8.0 through the current development tip (PostgreSQL 18) using PGBench from PostgreSQL 18, TPCH, and mid‑2010s hardware; OLTP scalability improves dramatically (typical 20–50x on many cores, prepared statements reaching ~300k tps in some cases) while single‑client throughput changes little; bulk load time for 10GB drops by ~50% between early releases and modern ones with parallel index builds in 11 helpful for maintenance; profile‑guided tricks using bolt profile‑guided optimization show up to 30% improvement on OLTP but only minor OLAP gains, and true OLAP progress likely requires columnar storage.
AI-written summary · about
chapters (10)
- 0:00Introduction to talk
- 1:46Why performance archaeology
- 8:58OLTP benchmark setup
- 17:58OLTP results and scalability
- 24:26OLTP improvements and optimizations
- 29:14OLAP benchmark setup
- 33:37OLAP query performance
- 42:22OLAP improvements and future
- 45:16Summary and future outlook
- 48:29Q&A and closing
related talks
29mPerformance Archaeology - 20 years of improvements
56mPerformance Improvements in PostgreSQL9 2
39mPostgreSQL BI Performance & News from XL project
31mDemanding the impossible rigorous database benchm
53mVisualizing Postgres I O Performance for Development
36mBest Practices in PostgreSQL Tuning: Navigating Key Performance Bottlenecks in the Cloud
36mAutomated Performance Testing of Postgres
31mPostgres Performance for Humans