Orchestration vs Choreography, A Guide To Composing Your Monolith
summary
Argues orchestration wins when workflows are complex or need coordinated rollback, while choreography scales and improves independence; recommends an orchestrator per workflow to isolate failures. Calls out specific tools and patterns: Kafka (durable log), AMQP exchanges, AWS Step Functions, Pact, OpenAPI/AsyncAPI, Avro/Thrift and OpenTelemetry for observability. Cites Lehman’s laws about quality and decay, a Sky Bet outage where a ship’s anchor severs three of four links, and a dependency board that drove delivery to near stall. Prefers events over commands but warns of infinite loops, schema evolution and control-plane needs.
AI-written summary · about
chapters (9)
- 0:00Orchestration vs choreography
- 3:09Software architecture for change
- 6:40E-commerce workflows
- 9:07Orchestration deep dive
- 15:04Choreography deep dive
- 22:48Choosing between orchestration and choreography
- 27:41Dangers of coupling
- 33:35Key takeaways
- 40:59Q&A
related talks
38mChoreography vs Orchestration in long running asynchronous processes
59mOrchestration vs. Choreography: The good, the bad, and the trade-offs
01h 13mOrchestration vs. Choreography in Microservices
32mBalancing Choreography & Orchestration
33mOrchestrate Your Choreography
01h 01mOrchestration vs. Choreography: The good, the bad, and the trade-offs
50mChoreography vs Orchestration in Serverless Microservices
46mChoreography vs Orchestration in serverless microservices