Index

Module 02 · Data SystemsData on many machines

When Machines Fail

When Machines Fail
When Machines Fail · Open board

The moment your program runs on more than one machine, three things you took for granted about a single computer stop being true. This lesson shows each one breaking.

Before you start. The previous lesson, Partitioning, cut one dataset into pieces and gave each machine only its piece. Earlier, Transactions gave you one machine’s promise: a transaction finishes whole or leaves no trace, because the one machine running it is always there to keep it. This lesson asks what happens to that promise once the one machine becomes two, like the leader and follower from Replication, talking over a wire that can be cut, delayed, or go quiet.

Continue reading with a plan

The first three lessons of every track are free. This lesson is included with a plan, which unlocks every lesson, board, video, lab and paper for one price.

View plans →Already a member? Sign in

Lab

A follower sends a heartbeat every second and the leader gives up after some silence. How much silence is enough?

Drag How much silence before giving up and watch Live followers called dead fall while Seconds to notice the crash climbs.

Open The heartbeat timeout →

Related paper

Review Data Systems →

Saved in this browser, and to your account when you sign in.