Pushing past PostgreSQL’s vertical scaling limits - Running Amazon RDS

on YouTube ↗40m

" summary

Amazon RDS runs the core metadata Postgres instances at hundreds of thousands TPS and hits vertical limits, so teams use sharding, read replicas, caching, partitioning and autovacuum plus tools like RDS Performance Insights and pg_stat_statements to find hot queries. Major upgrades use the RDS snapshot (instacart) logical replication flow with frozen validation replicas and a dedicated dump replica for maintenance. Operational fixes include connection limits, circuit breakers, simple SQLs, timed query kills, and a surprising predicted 10–12% CPU win that yielded a 133% CPU decrease after upgrade.

AI-written summary · about

" chapters (6)

1. 0:00Introduction and background
2. 8:33Performance management overview
3. 19:53Major version upgrades
4. 26:23Connection and query management
5. 33:51Wish list for PostgreSQL
6. 38:26Q&A and closing

" related talks

• 19mPostgreSQL at AmazonJim Mlodgenski · Postgres Conference Orlando 2025
• 53mPostgres in Amazon RDSDenish Patel · PostgresOpen 2014
• 41mRunning Highly Available Postgres in AWS RDSSushant Mahesh Pawar · FOSSASIA Summit 2017
• 54mBreakout SessionGrant McAlister · PGConf Silicon Valley 2015
• 18mPostgreSQL at AmazonJim Mlodgenski · PGConf India 2025
• 51mScaling Postgres to the next level at OpenAIBohan Zhang · PGConf.dev 2025
• 46mHow we made PG Fitter, Happier, More ProductiveAndrew Atkinson · PGConf NYC 2021
• 25mData-intensive PostgreSQL: Three ways to scaleMarco Slot · POSETTE: An Event for Postgres 2024