Panel: Microservices, Where Did It All Go Wrong
summary
Argues microservices work when chosen for specific needs, not as a default, and shows common failures: rushed boundary decisions during hypergrowth, brittle inter-service coupling, and runaway operational costs — one bank keeps 2,000 microservices and an audit-log system far larger than production. Recommends simulating future stresses and using residuality theory to find the Goldilocks service count. Discusses practical tech: event logs for auditing, data mesh versus single warehouses, and when to use Dynamo-like key value stores versus relational systems.
AI-written summary · about
chapters (7)
- 0:00Intro
- 0:38Were microservices a good idea?
- 12:06What have we learned from microservices?
- 15:14Are microservices issues technical or organizational?
- 27:05How to feasibly build a simulation of a system?
- 33:55Difference between transactional & analytical data
- 42:06Outro
related talks
55mMicroservices, Where Did It All Go Wrong
01h 00mMicroservices, Where Did It All Go Wrong?
01h 05mPanel: Were Microservices a Huge Mistake?
38mMicroservices, a bittersweet symphony
54mThe Long Sad History of Microservices
57mEmbracing Microservices
43mHow to Avoid Doing Microservices Completely Wrong
23mWhy You Had a Bad Time With Microservices