Lightning Talks on Distributed Systems
summary
Leader election uses a five-node quorum where a majority (3) picks a navigator and leaders send heartbeats so partitions let others elect a new leader; Paxos implements consensus with proposers, acceptors and learners, unique ordered proposal numbers and a Paxos prepare/accept two-phase that ensures one chosen value; Cassandra uses Paxos for writes; Google Cloud Spanner combines Paxos, leader election, redundant network and nodes, embedded retries and TrueTime bounded clock uncertainty to get global ACID with high availability most of the time.
AI-written summary · about
chapters (8)
- 0:00Introduction to talk
- 1:44Leader follower in clusters
- 4:57Paxos consensus protocol
- 11:05CAP theorem explained
- 13:34Consistency in distributed systems
- 17:08Availability and Cloud Spanner
- 20:38Magic of distributed systems
- 21:02Closing remarks and diversity
related talks
40mConsensus Algorithms in Distributed Systems
48mDistributed Consensus Algorithms
43mAll Together Now: An Introduction to Distributed Consensus
22mLiberating Distributed Consensus
51mDo Not Reimplement Distributed Consensus Algorithms
40mImplementing Distributed Consensus
41mDistributed Systems explained with NodeJS
16mSwiftPaxos: Fast Geo-Replicated State Machines