CSE Roadmap

Common tier · about 3 hours · 6 parts

Part 1 of 6 · 1 min read

Why this page follows the one on faults

In this module you will

  • Explain why no failure detector is perfect and how pings and heartbeats work.
  • State the two properties a leader election must have: safety and liveness.
  • Walk through a Raft election with terms, votes and a random timeout.
  • Build a simple election from a store with compare-and-swap and a time to live.
  • List the costs of having a leader and ways to limit them.

Module 4 showed that a node cannot tell a slow peer from a dead one. Yet many systems need exactly that call: which server should stop getting traffic, and which follower should take over when the leader goes quiet? This page shows the practical answer. It draws on the book "Understanding Distributed Systems", with links to the fencing tokens of module 4. The numbers are made up to show the idea, unless the text names the book.