Site Reliability in the Serverless Age
summary
Defines serverless by concrete tradeoffs: AWS Lambda and other managed services reduce ops but introduce hidden limits and costs, for example Lambda has no SLA, concurrent execution quotas like 500 concurrent invokes in us‑east and surprise bill spikes from recursive S3→Lambda→DynamoDB loops. Advocates treating cost as a first‑class metric, building SLOs from provider SLAs (or compensating for their absence), and investing in observability that includes automated analysis not just logging to avoid catastrophic scale and cost failures.
AI-written summary · about
chapters (9)
- 0:00Start
- 2:56Serverless is a Spectrum (AWS edition)
- 12:03Deeper Dive on the hard stuff
- 13:07Track your complete system cost
- 13:58Observability vs. Monitoring
- 15:08FYI: This isn't Analysis
- 15:56Serverless Capacity Planning Know thy limits
- 17:26Thinking about Cost and Architecture
- 19:36Prediction - A New DevOps Tribe: Fin DevOps
related talks
24mFinDevOps and Site Reliability in the Serverless Age
55mConfusion In The Land Of The Serverless
51mConfusion In The Land Of The Serverless
44mServerless Architecture on Production: Good, Bad and Ugly Parts
58mServerless - the brief past, the bewildering present, and the beautiful (?) future
01h 00mServerless - reality or BS - notes from the trenches
52mConfusion In The Land Of The Serverless
28mServerless - the way microservices should be