Distributed MySQL Architectures: Past, Present and Future
summary
A practical survey of MySQL at scale: single instances hit >1M QPS on simple hardware and handle tens of thousands of updates, yet large apps force distribution and much higher complexity. Historic hacks include up to 150 replicas and brittle ring replication; NDB Cluster promises transparent sharding but proves hard to operate. Modern tooling that actually solves problems includes Vitess (used by YouTube / now PlanetScale), ProxySQL for routing, Orchestrator for automated failover, and MySQL 8 features—semi‑sync, parallel replication, row‑based replication, atomic DDLs—that materially reduce data loss and operational pain.
AI-written summary · about
chapters (13)
- 0:00Intro
- 0:36Cover Some Basics Review History of Distributed MySQL Architectures Look at Current State of Technology And what Future is likely to bring us
- 1:17MySQL Deployment on More than one System within same Environment
- 3:10Distributed Systems Tend to be More complicated to Develop Against More Complicated to operate Have Additional Performance Bottlenecks Have Complicated Failure Modes
- 5:01Propagation Asynchronous
- 6:00Scaling Data Size
- 10:19Other Problems Operating at Scale Asynchronous Replication Replication which is not Crash Safe Non-Atomic DOL Single Threaded Replication Statement Based Replication Automatic Failover
- 11:57Query Routing is the first problem people think about Multi-Shard Queries Shard Re-Balancing (online)
- 14:02Mostly In Memory Technology with Automatic Sharding On Paper solves a lot of Problems Very Limited Traction in Practice
- 15:57Many Organizations Invested in their own Sharding Implementation Technologies
- 17:04Vitess is leader in MySQL Sharding Space Simple Sharding can be done with ProxySQL You can find many project of unknown quality on GitHub
- 22:24Theoretical Scalability Beats Efficiency at the early stage
- 25:30How Many Solutions will be Open Source vs only Available from Specific Cloud ?
related talks
24mComparing the High Availability of MySQL with Modern Distributed SQL Databases
52mTo Shard or Not to Shard
28mThe Next Decade in Data Management
41mThis is Not a Web App: The Evolution of a MySQL Deployment at Google
55mThe Future of MySQL
47mApplication Architecture Using MySQL
25mData in the Cloud
35mState of the Dolphin