Building Modern Microservices
summary
Describes why a production outage happens when services block on a slow backend (thread starvation diagnosed via Splunk) and contrasts anti-patterns like the distributed monolith with true microservices using database per service and event-driven messaging. Explains queues vs topics, Kafka/Kinesis tradeoffs, the atomic need to write local state then publish an event, and how the saga pattern uses compensating transactions for eventual consistency. Ends with a serverless example using kinesis + lambda and practical tracing and read-model recommendations.
AI-written summary · about
chapters (9)
- 0:00Introduction and background
- 7:55Monolith vs microservices
- 11:09Call time coupling issues
- 15:24Eventing mechanisms overview
- 23:13Orchestration vs choreography
- 27:06Saga pattern explained
- 31:18CQRS and event sourcing
- 37:39Serverless microservices
- 39:52Conclusions and key takeaways
related talks
01h 03mMicroservices vs The Distributed Monolith
37mBeyond Microservices: Streams, State & Scalability
50mEmbracing Microservices
44mMicroservices, Where Did It All Go Wrong
43mPanel: Microservices, Where Did It All Go Wrong
49mChallenges in Implementing Microservices
50mChallenges in Implementing Microservices
47mEvent-Driven Microservices = Microservices Done Right