For each lecture, we'll post slides, an outline of the content, a recording, and supplemental reading. Slides will be up before lecture (but not necessarily at 9:00am); we post slides with all animations (good for following along exactly) and with limited animations (good for studying/taking notes on). The outline and recording will be posted after lecture (but not necessarily at 3:00pm; likely the next day or so). You can do the supplemental reading before or after lecture; whatever works best for you.

• Slides: all animations, limited animations
• Outline
• Recording
• No supplemental reading for this lecture

After this lecture, at a minimum, you should be able to do the following:

• Explain the problem with simply extending 2PC to multiple servers and multiple coordinators
• Explain the problem that a network partition causes when we use a primary/backup mechanism
• Explain the role of every component in a replicated state machine (RSM): coordinators, the view server, and primary and backup servers.
  • This includes any data structures they keep, how they respond to requests, whether they communicate with other components, etc.
• Explain the role of every component in an RSM when there are failures (server failures, network partitions).
  • This includes how a component would discover a failure, any actions they need to take in response to the failure, what would happen while the failure is being handled, etc.
• Explain why we allow for the view server to be a centralized component, but did not decide to centralize the role of the coordinator
• After Recitation 20: Put the problem that Raft solves in the context of RSMs

Practice problems

• 2019, Problem 3 (Solutions)
• 2017, Problem 6 (Solutions)
• 2016, Problem 11 (Solutions)