DBtune: AI-driven performance tuning for all PostgreSQL flavors
summary
DBtune v3 autotelic agent installs in ~10–15 minutes and requires pg_stat_statements required to collect queries and metrics; it edits only postgresql.conf (revertible) and supports community Postgres plus RDS, Aurora and Google Cloud SQL. An open-source Go agent runs iterations (default 5‑minute windows) while an online ML with ~30 iterations retrains per instance, tuning 18 server parameters in the demo. Typical runs finish in 2–3 hours, report up to 2–2.5x latency improvement, and claim per‑instance cost savings (example: ~$8k/year).
AI-written summary · about
chapters (11)
- 0:00Introduction to DBtune
- 10:37Why database tuning matters
- 16:36When to tune PostgreSQL
- 23:14How DBtune works
- 28:49DBtune safety and scalability
- 30:18Demo setup and agent installation
- 34:14Monitoring workload metrics
- 35:36Workload fingerprint and tuning
- 39:51Tuning results and improvements
- 41:36Version 3 features and roadmap
- 47:41Q&A session
related talks
27mBuilding safety tooling for risk-free AI tuning of Postgres
28mAutotuning PostgreSQL on Azure Flexible Server
20mDebate: Autotuning
04mPostgres Tuning is saved by Sovenath Shaw
44mContinuous Database Administration
18mWhy Postgres Is Better Than MySQL for Automated Knob Tuning
48mChatGPT Ain’t Got $%@& On Me! Next Generation Automated Database Tuning
18mPostgreSQL: Tuning parameters or Tuning Queries? with Henrietta Dombrovskaya