We're Marketers. If we can Learn Distributed Systems + K8s, so can you!
Talkat KubeCon + CloudNativeCon North America 2021 · on Friday, 2021-10-15· at 16:30· Petree Hall C + Online·
summary
Explains why Kubernetes chooses etcd + Raft for a CP design, compares implementations (Docker Swarm, HashiCorp Nomad, Mesos, K3s) and shows tradeoffs: K3s uses a single binary and can swap etcd for a SQL store; Nomad emphasizes pluggable service discovery; Swarm embeds clustering in the Docker engine. Covers Raft mechanics (leader, majority commit), read/write latencies, and a real cascading failure where Kafka disruptions trigger a thundering herd and ~41,000 nodes spin up, overwhelming service discovery. Suitable if implementation details and failure cases matter.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:02Failure is inevitable
- 2:23Not a new problem
- 3:20Need to answer new questions
- 5:04Example....
- 6:30Why is it so hard?
- 9:16WTF is RAFT?
- 10:11Majority rules
- 10:40Leader Gets a Write
- 11:00Leader stages to log
- 11:12Leader says to commit
- 11:27Heartbeat messages
- 11:49I'm the eaptain Leader, now
- 12:42Components
- 14:11Example of when things go wrong
- 16:28Timeline
- 17:34Modern Distributed Systems
- 19:58MESOS
- 23:07SWARM
- 27:21NOMAD
- 30:13Kubernetes
related talks
20mKeynote: Kubernetes: This Job is Too Hard: Building New Tools, Patterns and Paradigms to Democratize
36mThe Road to More Usable Kubernetes
54mClosing the Developer Experience Gap in the Kubernetes Ecosystem
34mIntroduction to Kubernetes for Developers
44mTop new CNCF projects to look out for
34mScaling Cloud Development Environments on K8s: Enterprise-Scale Adoption Lessons
19mPains and gains of Kubernetes for Developers, story from experience
52mWeek 3 Live Q&A