Distributed Systems
The theory you actually get asked about — consistency, consensus, ordering, and partial failure.
9 topics
CAP & PACELC
What the theorem actually says, why "CP or AP" is usually a sloppy way to describe a database, and the extension that matters more in practice.
core20 minConsistency Models
The vocabulary between "strong" and "eventual" — linearizability, causal consistency, and the session guarantees users actually notice.
core25 minReplication & Quorums
How copies of data stay in step — single-leader, multi-leader and leaderless — and the R + W > N arithmetic behind tunable consistency.
core25 minConsensus & Leader Election
How a group of machines agrees on one value or one leader — enough Raft to use it correctly, and the fencing detail that prevents split brain.
core25 minClocks & Ordering
Why wall-clock timestamps can't order events across machines, and what to use instead — logical clocks, vector clocks and monotonic IDs.
deep20 minIdempotency & Delivery Semantics
Why exactly-once is a marketing term, and how idempotency keys turn at-least-once delivery into a system that doesn't double-charge anyone.
core25 minDistributed Transactions & Sagas
What to do when one operation must change state in several services — why 2PC is usually the wrong answer, and how compensation works instead.
core25 minPartial Failure & Failure Detection
The defining property of a distributed system — some of it is broken and you can't tell which — and what to do given that you can never be sure.
core20 minIdempotency
Making an operation safe to repeat, which is the only way a retry over an unreliable network can ever be correct.
core25 minshared