!!! Do not publish or share !!!!
summary
Argues ten concrete claims: asynchronous messaging favors throughput over latency, queues invert control flow (pushers vs fetchers) and require explicit flow control like TTL, tail‑drop or backpressure; retry without exponential backoff + jitter retry amplifies failure and topples systems; ordering kills concurrency so most systems shard/order by key; exactly‑once is impractical so prefer idempotency over exactly‑once. Uses AWS examples (EventBridge vs Pipes, SNS/SQS, Lambda), notes SQS limits (20,000 in‑flight per message group) and a serverless coffee demo that exposes real operational tradeoffs.
AI-written summary · about
chapters (17)
- 0:00Introduction and welcome
- 1:07History of integration patterns
- 2:52From fallacies to truths
- 5:00Core messaging concepts
- 12:28First truth: Two boxes and a line
- 14:37Why messaging is powerful
- 28:49Understanding coupling
- 46:07Gregor’s law of distributed computing
- 57:24Integration vs distributed systems
- 1:04:07Control flow in messaging
- 1:38:42Break
- 1:39:24Resilience and robustness
- 1:53:40Retry logic pitfalls
- 2:05:12Order and duplicate messages
- 2:23:19Event-driven architectures
- 2:38:14Architectural trade-offs
- 2:41:55Q&A and closing
related talks
40mEverything Will Flow
25mAvoiding Overload in Distributed Systems
52mWhy today, more then ever you need Dependency Isolation in Distributed Systems
44mBuilding Distributed Applications with Event-driven Architecture
44mFrom Synchronous to Asynchronous Distributed Microservices
52mDIY Async Message Pump: Lessons from the trenches
46mIntegration patterns & trade-offs for loosely coupled systems
47mYou Keep Using That Word