Serverless Architecture on Production: Good, Bad and Ugly Parts
summary
Companies run production workloads on AWS Lambda with concrete tradeoffs: a fully serverless incident management system is built and runs but requires heavy custom monitoring; a 220-line function replicates DynamoDB across regions and scales automatically (220-line cross-region sync Lambda). Cold starts and the memory/CPU tradeoff matter (128 MB runs ~12s vs 1 GB ~1.5s for similar cost). Operational surprises include an $4,000 billing incident from an infinite retry loop and hitting the 3,000 default concurrency limit.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 6:34Why serverless matters
- 11:57AWS Lambda deep dive
- 19:57Our serverless journey
- 25:00Key takeaways and challenges
- 37:24Use cases and best practices
- 42:19Final thoughts and Q&A
related talks
40m4 Key takeaways from running Serverless on production for 4 years
47mServerless in production, an experience report
29m2 Years, 0 Servers: Lessons learned from running a startup 100% on Serverless
01h 02mServerless in production, an experience report
39mServerless in production, an experience report
54mServerless beyond Functions
17mHow do we use Serverless Architectures on AWS Lambda
05m#Serverless + #Devops = BFF