Avoiding Microservice Megadisaster
summary
Post-mortem of an 18‑month rewrite that spends $150M and hundreds of developers to replace a monolith and ends up with 200+ networked API calls that produce a nine and a half minute page load. Diagnosis: synchronous chains, mistaken SLAs, and local devs never crossing network boundaries. Fix implemented for search: feed a document DB denormalization pipeline via nightly batches and DB triggers so the UI hits one fast store. Worked for search, not for the whole site.
AI-written summary · about
chapters (19)
- 0:00Avoiding Microservice Megadisasters
- 0:33Bell Computers
- 5:00Netflix-oriented architecture
- 5:32Core principles
- 6:52Pre-production go-live
- 11:21API hell
- 12:13Doing the math
- 16:04How Not to Do It (MS Style)
- 17:33Traditional n-tier
- 17:48Areas of Responsibility
- 20:29what is a Service? Software that is
- 21:15what is a Microservice?
- 21:44what is a Bounded context?
- 25:00Grounding in reality
- 28:38Data ownership
- 31:05Catalog
- 33:39Pricing
- 35:18Localization
- 36:45Content
related talks
51mAvoiding Microservice Megadisasters
45mAvoiding Microservice Megadisasters
01h 01mMicroservices Gone Wrong
48mMicroMonolith
22mMicroservices Highway to hell
48mExtracting a microservice from a 15-year-old monolith: a one-year adventure
42mDo proper monoliths before you do microservices!
42mArchitecting for Failures in Microservices