Lessons learned scaling our SaaS on Postgres to 8+ billion events
summary
ConvertFlow scales from Rails on Heroku Postgres to a Citus sharded cluster on Azure, processing 8.7 billion events and serving ~40M monthly visitors; switching off Firebase cuts cloud costs. Shard analytics by tenant ID with Citus (Hyperscale) to get 2–4x faster queries, use rollup tables and daily increments to reduce query work, reindex CONCURRENTLY to avoid locks, and perform a bigint migration for primary keys with a short maintenance window.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:58What is Convertflow
- 2:27From 0 to 1M events
- 6:42From 1M to 100M events
- 8:57From 100M to 1B events
- 12:13From 1B to multiple billion events
- 15:03Bigint migration
- 16:08Migration to Hyperscale (Citus)
- 17:19Lessons learned recap
related talks
52mScaling a Saas application beyond a single Postgres with Citus
24mCitus from the Customer eyes
16mOnline Scale-Out and Cloud Data Ingestion
24mLessons learned: Migrating from AWS-Hosted PostgreSQL RDS to Self-Hosted Citus
55mBreakout Session
32mAuto scaling Azure Cosmos DB for PostgreSQL with Citus, Grafana & Azure Serverless
02h 02mAzure Cosmos DB: Liftoff
16mKeynote – Umur Cubukcu
