So You Want to Geo-Partition Your Data... - Deepthi Sigireddi & Jiten Vaidya, Planetscale, Inc.
summary
A production Vitess cluster runs across four GKE clusters (Asia, Australia, Europe, US) with a sharded user DB keyspace split into four shards; a custom Vindex (40–50 lines of Go) maps country → region by writing a one-byte region tag into the computed keyspace_id so inserts with only ID, country, name land in the correct shard. The demo shows scatter‑gather reads and that rows appear only in their jurisdictional shards, and cross‑shard operations use a two‑phase commit with redo logs for safe failover.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:37Problem
- 3:47What is Witness
- 9:08Cell
- 17:01Cluster
- 18:54V Schema
- 20:53User DB Key Space
- 21:32User DB Data
- 23:29Mapping Table
- 24:42Question
related talks
48mGeo-partitioning with Vitess
34mSo You Want to Geo-Partition Your Data...
38mHow Geo-Distributed Apps Use Global Cloud Infra for Speed and Compliance
34mInside Vitess: Building Resilient, Distributed Databases
35mExploring High Availability in Kubernetes with Vitess
45mDeep Dive: Vitess
31mToo Far Too Furious: Multi-Region Cloud Applications
37mSharding: Growing Systems from Node-scale to Planet-scale