Scaling out by distributing and replicating data in Postgres XC
summary
Postgres-XC is a shared‑nothing PostgreSQL cluster with coordinators, datanodes and a GTM; the team builds a cluster, adapts table creation and runs DBT‑V schema benchmarks to prove concrete effects of layout. Replicate tables for high point‑read load, hash/round‑robin distribute for write scaling, and pick distribution columns that match equality joins to push joins to datanodes. Wrong layout drops throughput dramatically — 2500 TPS vs 250 TPS — and adding a GTM proxy speeds 2–3x; global constraints and dynamic rebalancing remain work in progress.
AI-written summary · about
chapters (9)
- 0:00Introduction to Postgres XC
- 6:58Data distribution strategies
- 15:21Query processing in Postgres XC
- 28:22Choosing distribution strategy
- 33:43Load balancing and constraints
- 40:38DBT2 benchmark results
- 46:39Scaling tips and GTM proxy
- 49:55Fault tolerance and recovery
- 53:12Q&A and resources
related talks
50mSupersized Postgres: Postgres-XL for OLTP and Big Data Analytics
58mPostgres-XC dynamic cluster managment
47mWhat Is Postgres XC?
59mUsing Postgres XL
39mPostgreSQL BI Performance & News from XL project
41mRevisiting XTM: A Practical Case Study Highlighting Its Needs
25mData-intensive PostgreSQL: Three ways to scale
48mHow to Tame A Mastodon: Lessons for PostgreSQL at Scale