Change your Architecture During Deployment
summary
Claims that changing the physical deployment can alter dependencies and trade offs: Prime Video allegedly moved from distributed lambdas back to a single process because of a hard limit around 5x expected load; a demo uses ASP.NET view model composition and a composition gateway that uses reflection to load ICompositionRequest handlers from deployed assemblies, bundling frontend page pieces and backend calls into the same process. That reduces network RPCs, exposes connection strings at deploy time, and in the demo yields fewer remote calls and lower latency — a concrete technique that actually runs.
AI-written summary · about
chapters (14)
- 0:00Changing architecture during deployment
- 3:24Opinions vs facts in architecture
- 5:10Communicating architecture with C4
- 9:13Microservices myths and maintainability
- 12:07Data ownership in architecture
- 15:45Shared databases impact
- 17:27Prime Video’s monolith shift
- 19:18Microservices oversold by vendors
- 20:49Data ownership in UI composition
- 24:09Fixing API complexity with composition
- 28:47Composition gateway demo
- 32:53Balancing architecture boundaries
- 34:00Key takeaways on data ownership
- 35:21Q&A
related talks
01h 29mMonolithic Delivery
41mSurviving Continuous Deployment in Distributed Systems
08mHow Your Architecture & Infrastructure Can Make or Break Your Team’s Productivity
41mASP.NET Core in Production: Lessons Learned
26mArchitecting for scale
53mChange your tools, change your outcome
50mHow to be cool in the age of legacy
01h 01mFrom one release per quarter to 30 times a day