PastgreSQL: Dead data tells all tales!
summary
A Postgres auditing system called audit_star stores row images as a single JSON column generated by row triggers, deploys via a one‑line Go tool across hundreds of clusters, and exposes three views (Delta/Snapshot/Compare) for stitched queries. Benchmarks with pgbench TPC‑B and ~30 trigger variants show typical overhead ≈30% TPS; B‑tree index adds ~10% and dynamic SQL ~30%. Design choices (capture only deltas, prune/archive, avoid heavy JSONB indexing) keep disk growth small and the system viable in production.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:07Innova International
- 2:02Agenda
- 2:56Why did we build it
- 7:23Deployment Utility
- 8:08What could go wrong
- 9:40Key benefits
- 10:53Design considerations
- 16:22Testing
- 22:50Overhead
- 24:21Demo
- 30:12Comparing Snapshots
- 36:15The Pattern
- 41:53GitHub
- 42:25Future optimizations
