Performance is a feature: accelerating Postgres
summary
Shows that moving compute into Postgres via extensions and PL languages yields big wins: a moving‑average job on 1M transactions runs ~1.8s locally vs ~12s over the network, so network is a huge factor but Python interpreter overhead also dominates. Uses PGX to build PL/Rust extensions, compares PL/Python vs PL/Rust, and converts perf.data for perf → Firefox profiler visualization to pinpoint hotspots. Result: Rust extensions run closer to C, safer for managed RDS, and benchmarks validate the approach.
AI-written summary · about
chapters (7)
- 0:00Introduction and goals
- 5:05Postgres as a platform
- 11:10Postgres extensions and PL Rust
- 15:04Benchmarking PL Python vs PL Rust
- 21:51Performance as a feature
- 27:32Database as an operating system
- 38:05Q&A session
related talks
36mAutomated Performance Testing of Postgres
32mHacking Postgres Executor For Performance
46mFearless Extension Development With Rust and PGRX
56mLightning Talks
50mPerformance Archaeology
36mBest Practices in PostgreSQL Tuning: Navigating Key Performance Bottlenecks in the Cloud
31mPostgres Performance for Humans
34mPostgres 10, Performance, and You