Is there anybody out there
summary
Crash course on building reliable clusters that names concrete tools, tradeoffs and numbers: group membership hinges on failure detectors (uses phi accrual and a newer adaptive accrual implemented and benchmarked by the presenter), dissemination relies on gossip and piggybacking (SWIM style extended by SWIM + Memberlist (<2% false positives)), and consensus uses Lamport clocks, replicated state machines and algorithms like Paxos/Raft and the new CASPaxos (logless consensus). Akka Cluster combines ping‑pong failure checks, biased random gossip and leader‑by‑convention membership decisions.
AI-written summary · about
chapters (32)
- 0:00Introduction
- 0:23A career in distributed systems
- 1:10A selfless act
- 1:42About me
- 2:26Motivational quote
- 3:37Distributed systems
- 4:13Clusters
- 5:38Three things
- 5:58Wish You Were Here
- 6:23Failure Detectors
- 7:28Impossible Results
- 8:49Failure Detection
- 12:37Swim
- 13:55Indirect probing
- 15:09Hoshi Corp
- 16:31Pulse
- 17:03Multicast
- 17:32UDP multicast
- 18:21Napster
- 20:50Infection Stag Gossip
- 21:48Fail Detection Gossip
- 23:23Gossip Optimizations
- 25:11Dissemination
- 25:27Consensus
- 25:46Group Membership
- 27:41Time and Time
- 29:03Version Vectors
- 31:11State Machines
- 33:19Consensus Protocol
- 37:01Conventions
- 38:13A Cluster
- 40:27A Happy Path
related talks
46mFault tolerance in Distributed Systems: A Case Study in Apache Spark
22mAdvanced Akka features
37mBeyond Fault Tolerance
44mConcurrency in Scala and on the JVM
05mJacek Kunicki Interview
39mCloud Native Akka & Kubernetes The Holy Grail To Elasticity
46mPerfect Scalability
58mKeynote: Patterns In The Chaos