System design and architecture · Proficient · Lesson 64 of 98 · about 2 weeks
Distributed systems fundamentals: consistency, consensus and failure
After this you can explain the CAP trade-off, replication consistency models, and why networks and clocks make distributed systems hard.
Why it matters
The moment you have two machines, you have a distributed system, and things you took for granted stop being true: messages get lost or delayed, clocks disagree, and a node that looks dead may be alive. Most production incidents in larger systems come from these assumptions breaking. Kleppmann's Cambridge lectures are the best free treatment. Watch lectures 1 to 5 at least (failure models, time, replication, quorums). The CAP video gives you the interview-level version. Trap: treating CAP as a menu where you pick two. Partitions happen; the real choice is what you do during one.
How to use this resource
Prioritise lectures 1 to 5 (about four hours). Lectures 6 to 8 on consensus, Raft and linearizability are excellent if you work on infrastructure.