KEYNOTE: The Distributed PostgreSQL Problem & How Citus Solves it
summary
Citus is a Postgres extension that shards and replicates tables across PostgreSQL servers to get “Postgres performance at any scale,” with a decade of engineering and open source code on GitHub and a managed option via Azure Cosmos DB for PostgreSQL. It enforces co-location by shard key and replicated reference tables to keep joins local, shows GigaOM/HammerDB distributed SQL benchmark wins when modeled for locality, and admits gaps (schema-level sharding, snapshot isolation) under active development.
AI-written summary · about
chapters (11)
- 0:00Introducing Marco
- 2:01Marco’s talk begins: what is Citus?
- 2:56Why we need distributed PostgreSQL
- 7:44Data locality problem
- 12:54How Citus addresses data locality
- 15:12Distributed SQL benchmark results
- 15:57Distributed data modeling problem
- 18:53Versatility problem
- 19:12Citus features & gaps
- 20:09Engineering culture problem
- 22:57Q&A
related talks
08mFoundations of Distributed PostgreSQL with Citus
32m8 keys to the growth of Citus & Distributed PostgreSQL
44mDistributed Postgres with Citus
16mKeynote – Umur Cubukcu
31mSharding Without the Complexity
28mEfficiently distributing Postgres with Citus – How to choose the right shard key?
38mPostgreSQL as a distributed computing platform
52mBuilding Distributed OLTP with Azure Cosmos DB for PostgreSQL
