Do Distributed Systems they said. It'll be fun they said.
summary
Four years of building an event store and messaging platform reveal concrete fixes: used Raft for consensus and replication across nodes, idempotent commands and deduplication with command IDs plus compare-and-swap to avoid double side effects. Practical techniques include retries, circuit breakers and failover, deadlines and compensating actions, consistent hashing for sequential dispatch, a message bus for location transparency, streaming with backpressure and batching, layered storage and sharding; each technique is presented with when it works and its tradeoffs.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 3:08Message dispatching
- 6:22Messages not deliver
- 9:22Failover
- 11:10Methods delivered multiple times
- 16:24Duplication
- 19:35Design solutions
- 22:09Transport protocol
- 26:12Location transparency
- 28:18Performance
- 32:10impedance mismatch
- 35:25replicated data
- 39:51sharding
- 41:31summary
related talks
50mDo Distributed Systems they said. It'll be fun
30mDo Distributed Systems, they said. It'll be fun, they said
52mThe Computer Science behind a modern distributed data store
57mReliability Nirvana
25mGrowing Distributed Systems
33mWhy Are Distributed Systems So Hard?
47mDistributed Systems: When Democracy and Computer Science Come Together
30mDistributed system design