Tobias Petry :: Scaling Databases
summary
Starts with concrete failures: multimaster replication produces conflicts under concurrent writes so throughput drops; switching to read replicas fixes read scale but creates eventual‑consistency bugs where freshly written orders vanish from secondaries; synchronous replication stalls writes waiting on the slowest replica, while quorum writes trade latency for availability; true write scaling requires sharding or a distributed engine like citus or Planet scale and moving join logic into application; practical tips include cache‑based master routing, per‑table sharding options, and using TablePlus for exploration.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:32Why database talk
- 2:24Story mode
- 2:51The problem
- 3:56Everyone has different needs
- 4:26Optimizing first
- 6:21Multimaster replication
- 14:09Synchronous replication
- 14:52Why its not instant
- 16:15Performance
- 17:20Sharding
- 22:19Scaling
- 22:56Data issues
- 24:26Questions
related talks
28mPractical Approaches To Scaling Distributed Transactions
47mDatabase Resiliency
28mSmall Data: Databases in the real world
41mEverything you always wanted to know about highly available distributed databases
01h 00mDistributed transactions implementation trade-offs
24mDatabases beyond the tutorials
21mHow Safe is Asynchronous Master-Master Setup?
30mDatastore Axes: Choosing the Scalability Direction You Need