Terrible Ideas In Lambda
summary
Calls out concrete limits and outcomes: Lambda fails for big stateful bundles above 75 MB, times out at 500 seconds, and can produce explosive concurrency (one experiment plans to hit 4.2 billion IPv4 addresses and estimates about $9,000). Shows real builds that do and don’t work: an efficient frame-by-frame ML job that costs ~$8 and finishes in five minutes, and brittle glue architectures using S3, SQS, SNS, DynamoDB, KMS and CloudWatch that create infinite-loops, race conditions, or support‑API storms.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 3:54When not to use Lambda
- 8:14Lambda infrastructure
- 12:55Terrible Lambda use cases
- 18:16Lambda certifications and SLAs
- 19:00Lambda triggers and risks
- 21:08Q&A
related talks
25mHow to Fail at Serverless (without even trying)
54mServerless beyond Functions
50mServerless Architecture Patterns and Best Practices
53mBottlenecks in Serverless Applications
32mUsing Serverless Functions for Real-time Observability
44mServerless Architecture on Production: Good, Bad and Ugly Parts
54mContainerless in the cloud with AWS Lambda
53mLook Ma! No servers!