Postgres Performance in 15 Minutes
summary
Claim: configuration tweaks account for less than 20% of performance while a handful of bad requests consume most resources; concrete fixes include removing unnecessary queries, adding indexes, fixing filter expressions, and putting large objects in a CDN. Tools and numbers: run PG Badger on full logs to find the few hot queries, use PG bouncer transaction pooling and read replicas for scale, and size IO/RAM (shared_buffers, work_mem) so Postgres reaches thousands of requests/sec.
AI-written summary · about
chapters (7)
- 0:00Start
- 5:36data you don't need
- 10:23transaction pitfalls bad: bundling reads into a transaction
- 11:27locking check for waiting queries
- 15:32the easy stuff use the latest version of PostgreSQL perf improvements in every release
- 15:54use pgbouncer for connection pooling
- 17:14load balancing special workloads
related talks
31mPostgres Performance for Humans
43mPostgres Performance: From Slow to Pro
45mProper PostgreSQL Parameters to Prevent Poor Performance
34mTuning PostgreSQL to work even better
47mIdentifying Slow Queries and Fixing Them!
50mImproving Postgres' Efficiency
50mPostgreSQL Query Performance Insights
46mHow we made PG Fitter, Happier, More Productive