Distributed Systems: When Democracy and Computer Science Come Together
summary
Explains how Raft provides consensus for RavenDB clusters using leader election, log replication and quorum commits, with 150–300 ms randomized timeouts and commit only after a majority so a five-node cluster tolerates two failures and a three-node cluster is risky. Uses simple RPC over HTTP/TCP, shows that uncommitted entries get discarded when a higher-term leader wins, and demonstrates practical failover and partition behavior—safety holds with a majority, partitions cause temporary unavailability. Raft chosen for understandability
AI-written summary · about
chapters (19)
- 0:00Introduction
- 0:54What is a distributed system
- 2:01Three main characteristics
- 3:17Challenges
- 4:06Cap Theorem
- 7:14Consensus
- 8:55Followers
- 11:36Terms
- 13:27Leader Election
- 13:57Vote Request
- 15:52Split Vote
- 21:54Leader
- 25:59Our Job
- 28:50No 3 Node Cluster
- 29:03Questions
- 29:48Election
- 31:48Meeting the Leader
- 39:33RavenDB
- 43:38Communication
related talks
47mUnderstanding Raft Consensus
41mThe Raft Consensus Algorithm
14mDistributed Consensus in 15 Minutes!
51mDo Not Reimplement Distributed Consensus Algorithms
52mThe Computer Science behind a modern distributed data store
44mCluster Consensus: When Aeron Met Raft
50mDo Distributed Systems they said. It'll be fun
01h 32mHow to Reach Consensus with Paper, Pencil and the Raft Algorithm