Kill "Microservices" before its too late
summary
Argues that the word "microservices" is useless and the goal is decoupling for change, describing Wunderlist as ~100+ tiny services talking over HTTP+JSON, enforced by immutable infrastructure (Chef images replaced, rebuild not mutate). Teams accept heterogeneous languages (Scala Haskell Clojure Go Ruby Elixir), forbid shared libraries to avoid coupling, and replace components often—80% of code gets rewritten while the system endures. Emphasizes production metrics and fast rollbacks instead of heavyweight unit-test permanence; one legacy Haskell service that never changed causes a painful outage.
AI-written summary · about
chapters (23)
- 0:00Intro
- 3:00Kill "Microservices" before it's too late
- 5:57Impermanance the ironic key to systems that survive
- 9:45Homeostasis
- 11:39Tiny components
- 12:46Massive, elite team
- 12:56Enormous, agile company
- 14:07Immutable infrastructure
- 16:21Disposable components
- 17:34Mutability of the system is enhanced by immutability of its components
- 18:08The system is the asset Code is a liability
- 20:11Radically heterogenous
- 21:48No shared code
- 23:52Code that works and has never been modified is good code.
- 26:10Immutable == disposable
- 26:24Metrics vs Tests
- 28:43Impermanent Organizations?
- 31:22"small" project == "impermanent" project
- 31:37Language stabilization
- 33:39Microservices is not an architecture.
- 34:17Healthy, longevity is the goal.
- 34:50Impermanence - Decoupling, Longevity, Agility
- 39:35Recording and Production Confreaks
related talks
55mMicroservices, Where Did It All Go Wrong
44mIs It Time Already? Microservices Are Dead; Long Live Microservices!
57mEmbracing Microservices
57mBuilding Systems that are #neverdone
01h 00mMicroservices, Where Did It All Go Wrong?
38mMicroservices, a bittersweet symphony
42mMicroservices are Not Worth the Trouble!!??
44mMicroservices, Where Did It All Go Wrong