Adventures in Spacetime
summary
Explains that ordering in distributed systems is fundamentally partial and shows the actual Lamport algorithm: keep a counter, increment on send, include timestamp, on receive set local = max(local, received) + 1 so causality is preserved. Demonstrates CAP as an impossibility — you cannot guarantee consistency, availability and partition tolerance simultaneously — and traces why common assumptions fail (network isn’t reliable, latency nonzero). Uses concrete incidents like the O2 outage (≈30 million users, expired certificate) and a network refractive index metaphor to show software adds measurable delay and failures.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 7:51Distributed systems and causality
- 15:39Distributed systems challenges
- 26:12Eight fallacies of distributed computing
- 37:15Distributed objects pitfalls
- 43:42CAP theorem implications
- 51:09PACELC trade-offs
- 56:28Refractive index and network speed
- 1:02:31Conclusion and Q&A
related talks
57mScaling the CAP Theorem for Modern Distributed Systems
56mOrdering the chaos - cleaning logs and ordering events in microservices
41mCloud Programming: CALM Lessons and a new PACT
45m(...)Distributed Systems
44mThe Road to Akka Cluster, and Beyond
34mThe Holy Grail of Systems Analysis: from What to Where to Why
25mThe shell game called eventual consistency
43mDistributed Systems