Paxos is the classic consensus algorithm for crash-fault-tolerant distributed systems.

It solves a narrow but important problem: a group of nodes must agree on one value, even when some nodes crash and messages are delayed, duplicated, or delivered out of order.

Paxos matters because its ideas appear in many later systems:

The algorithm has a reputation for being hard to learn. That reputation is earned. The basic protocol is small, but the safety argument is subtle, and real implementations need leader election, recovery, reconfiguration, snapshots, batching, and careful disk persistence.

This chapter focuses on the core protocol first. Multi-Paxos and production concerns come after the single-value version is clear.