Postgres Tuning is saved by Sovenath Shaw
summary
A tiny runtime agent of roughly 20 lines of code continuously gathers metrics and tests combinations of parameters like max_parallel_workers_per_gather and random_page_cost, detects the current workload shape, writes back suggested config changes and optionally applies them, avoiding one-by-one manual tuning. It chooses settings on a log-scale search to avoid brittle luck, measurably increases transaction throughput, and reports cost savings demonstrated on RDS m52 instances.
AI-written summary · about
related talks
18mPostgreSQL: Tuning parameters or Tuning Queries? with Henrietta Dombrovskaya
43mPostgres Performance: From Slow to Pro
48mDBtune: AI-driven performance tuning for all PostgreSQL flavors
52mPostgreSQL Configuration for Humans
18mTuning Parameters in Postgres vs. Tuning Your Queries
33mTuning PostgreSQL to work even better
34mTuning PostgreSQL to work even better
45mProper PostgreSQL Parameters to Prevent Poor Performance