Services, Services, Everywhere!
summary
Argues that splitting a monolith works only with concrete practices not slogans and shows what to build: whiteboard-driven service boundaries, one deployment system (Capistrano or equivalent), shared request-id tracing through logs, and client libraries autogenerated from HTTP/JSON contracts. Warns of real failures — N+1 over the network (many 50ms RPCs that wreck latency) and 10× deployment/config complexity — and gives fixes: denormalized documents, Futures/asynchronous calls, and brokers like SQS or RabbitMQ (avoid ZeroMQ until necessary) plus a monthly fire drill to verify logging and recovery.
AI-written summary · about
chapters (45)
- 0:00Intro
- 1:45Services.
- 2:48Small things?
- 3:34What size is micro
- 4:40Benefits
- 6:47Responsibility
- 7:49Discipline
- 9:17Velocity
- 10:41Boundaries
- 11:25Active Record chop
- 13:09The Breakup Redux
- 13:35Fix "touchers"
- 16:18Whiteboard Arch
- 17:46Deployment
- 18:30If a app is never deployed, was it written?
- 19:49Solutions
- 20:08Requirements
- 23:02Expect Failure
- 25:19Tracing
- 26:31How & What
- 28:57How - RPC
- 29:51RPC - Obvious
- 31:30Example: Finagle
- 32:04RPC Testing
- 32:35REST / Documents
- 33:19HTTP/JSON Agony
- 33:55Standardize
- 34:45Clients
- 36:11More Standardizing
- 37:03Message Passing
- 38:16Brokers
- 38:51What was I doing?
- 40:10Awkward, But Powerful
- 41:40Comm Delay
- 42:11Denormalized Docs
- 42:44DND
- 43:24Futures
- 45:23Has anyone seen my database?
- 46:20Multiple langs
- 46:49Code reusability
- 47:43Should I do services?
- 49:09When?
- 51:07Microservices?
- 51:57Do Your Homework
- 52:11Summary
related talks
40mService Architectures for Mere Mortals
01h 00mService Oriented Architecture for Robust & Scalable Systems
01h 03mMicroservices vs The Distributed Monolith
38mMicroservices, a bittersweet symphony
50mEmbracing Microservices
57mBuilding Systems that are #neverdone
45mContinuously delivering containerized, highly available microservices
36mDo Proper Monoliths before You Do Microservices!