Microservices, a bittersweet symphony
summary
Argues that microservices often introduce more operational complexity than they solve, backed by concrete rules and tradeoffs: use cognitive load to size services, require fast provisioning, centralized logs and unique request IDs, and be able to deploy a service in minutes. Cites a real extraction at Ride and a SoundCloud lesson about technology proliferation. Recommends bounded contexts, executable contracts, schema-based formats like Protocol Buffers or Capn Proto, prefers synchronous comms, warns that four 99.9% services yield ~96.1% system availability and invokes CAP tradeoffs.
AI-written summary · about
chapters (10)
- 0:00Start
- 0:59MICROSERVICES?
- 2:20MICROSERVICE BE?
- 2:57Team size
- 8:20Monolithic app
- 17:00heterogeneity
- 20:47deployment
- 24:25partitions
- 28:49theorem
- 34:06databases
