Citus from the Customer eyes
summary
Citus extends Postgres to scale from one node to many as an extension (not a fork) and is available managed as Azure Cosmos DB for PostgreSQL; production users report concrete results: Algolia runs two clusters (~20 nodes) with 10+ billion rows and sustains 100 qps with 95th percentile <800ms, HSL ingests >1B rows from 2000 buses and cuts query times from 10 minutes to instant while saving ~50% ops cost. Followable specifics: shard by a natural key, co-locate big tables and replicate small ones, use JSONB, time-partition shards, push predicates to the distribution key; migration tools include pg_dump/pg_restore, pgcopydb, pg_azure_storage, Azure Data Factory and built‑in managed features like PITR, rebalance, replicas and PgBouncer.
AI-written summary · about
chapters (5)
- 0:00Introduction
- 2:04What is Citus
- 3:33When to use Citus?
- 8:045 tips for distributed data-modeling with Postgres
- 16:34Running app in production with Citus
related talks
08mFoundations of Distributed PostgreSQL with Citus
16mOnline Scale-Out and Cloud Data Ingestion
02h 02mAzure Cosmos DB: Liftoff
52mBuilding Distributed OLTP with Azure Cosmos DB for PostgreSQL
32m8 keys to the growth of Citus & Distributed PostgreSQL
28mCitus Cluster: our zero-downtime migration story
30mBuilding geospatial apps on Postgres, PostGIS, & Citus at large scale
44mMulti-tenant SaaS apps made simple on Azure Cosmos DB for PostgreSQL