Distributed systems and scalability rules
summary
Compact overview of how real systems trade consistency, availability and partition tolerance with actionable patterns: use replication: master/slave vs multimaster and choose sync vs async depending on latency vs strong consistency needs, shard by a chosen key for scale, and prefer messaging and caching over cross-node locking. Mentions algorithms and limits: Paxos and CAP theorem, Byzantine failures require <1/3 faulty nodes as a hard bound. Recommends redundancy, data locality, avoiding single points and not using distributed transactions when possible.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:24Distributed systems
- 11:59Fundamental problems
- 17:21Key design principles
- 25:06Replication and sharding
- 35:19CAP theorem and approaches
- 40:16Design principles
- 43:24Q&A
related talks
43mWhat We Talk About When We Talk About Distributed Systems
22mWhy Are Distributed Systems so Hard?
41mIntroduction to distributed systems
25mGrowing Distributed Systems
25mThe shell game called eventual consistency
41mEverything you always wanted to know about highly available distributed databases
34mDistributed systems 102
28mPractical Approaches To Scaling Distributed Transactions