4 Key takeaways from running Serverless on production for 4 years
summary
Four years of running AWS Lambda at Opsgenie yields concrete lessons: production use hits concurrent execution limits and cold starts (Java + VPC used to double latency, AWS later fixed it), a runaway retry loop caused a two‑week bill spike (~$4,000) that required throttles and alerts, building an internal APM became Thundra and successfully monitors functions via CloudWatch logs, and small tools like auto-scale DynamoDB via Lambda cut ops work. Result: >99.999% availability, cheaper TCO for many workloads, but still needs strong engineers and proper CI/CD/monitoring.
AI-written summary · about
chapters (6)
- 0:00Introduction and welcome
- 3:08Why serverless matters now
- 12:26Serverless needs great engineers
- 16:45Serverless is less ops but not no ops
- 25:21Serverless is usually cheaper
- 37:15Wrap-up and key takeaways
related talks
44mServerless Architecture on Production: Good, Bad and Ugly Parts
17mHow do we use Serverless Architectures on AWS Lambda
29m2 Years, 0 Servers: Lessons learned from running a startup 100% on Serverless
47mServerless in production, an experience report
01h 02mServerless in production, an experience report
39mServerless in production, an experience report
29mGoing Serverless
05m#Serverless + #Devops = BFF