Five Common Mistakes With Distributed Systems
summary
Five common operational errors and concrete fixes: don’t split a system until you hit hardware limits because network latency and Gen2 GC causes stop-the-world pauses wreck throughput; don’t rewrite wholesale — rewrites drag on and create distributed monoliths; don’t build your own service bus — use NServiceBus or mature libraries instead of homegrown queues; use RabbitMQ TTL tricks or NServiceBus’s delay strategy (example: 28 queues = 8.5 years delay built to simulate long delays); solve some problems in policy not code (use negative inventory and reconcile later).
AI-written summary · about
chapters (8)
- 0:00Introduction and background
- 1:31Distributing too early
- 11:30Rewriting everything
- 16:09Building a service bus
- 26:29Defining services as nouns
- 32:49Striving for consistency
- 37:25Conflating boundaries
- 41:17Q&A
related talks
43mDistributed Systems
53mDon’t Build a Distributed Monolith: How to Avoid Doing Microservices Wrong
01h 04mDon’t Build a Distributed Monolith
43mHow to Avoid Doing Microservices Completely Wrong
41mTop 5 techniques for building the worst microservice system ever
33mWhy Are Distributed Systems So Hard?
01h 03mMicroservices vs The Distributed Monolith
48m9 Fallacies of Event driven Architectures