Do Distributed Systems they said. It'll be fun
summary
Axon engineers describe building a distributed event store and messaging platform using Raft for replication and consensus and a mix of concrete techniques: Raft consensus for durable commits, retries plus circuit breakers and failover for delivery, idempotency via idempotent commands or compare‑and‑swap and artificial message IDs to tolerate duplicates, consistent hashing to keep per‑key ordering, streaming with backpressure and batching for throughput, and layered storage and sharding for capacity. Practical tradeoffs and when each technique actually works are shown.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:39When to distribute systems
- 3:22Message dispatching challenges
- 6:55Messages not delivered
- 11:36Design solutions for delivery
- 13:03Messages delivered multiple times
- 19:30Messages delivered out of order
- 27:19Non-functional requirements changes
- 33:04Performance in distributed systems
- 38:11Data storage challenges
- 46:03Key takeaways and principles
related talks
45mDo Distributed Systems they said. It'll be fun they said.
30mDo Distributed Systems, they said. It'll be fun, they said
47mDistributed Systems: When Democracy and Computer Science Come Together
51mDo Not Reimplement Distributed Consensus Algorithms
52mThe Computer Science behind a modern distributed data store
40mConsensus Algorithms in Distributed Systems
44mCluster Consensus: When Aeron Met Raft
46mWhat can go wrong in a distributed system - experience from...