Dissecting serverless runtimes - from functions to databases
summary
Maps Lambda internals to other serverless runtimes: shows Docker/bootstrap loop that polls for requests, explains cold vs warm behavior with cold starts from milliseconds to hundreds and warm retention often minutes, exposes billing as gigabyte-seconds and a rough break‑even (~40 hours per month vs reserved VM), compares serverless Redis/Elasticache, CockroachDB and Redpanda pricing models, describes a Hazelcast prototype that restores data from S3, autoscaling via control plane, and recovery latencies typically 500–600 ms up to several seconds.
AI-written summary · about
chapters (15)
- 0:00Introduction and setup
- 1:31What is serverless
- 5:25How serverless functions work
- 9:18Serverless runtimes and Docker
- 12:09Cost model of serverless functions
- 16:04Cold starts and warm durations
- 20:25Complexity beyond single functions
- 21:54Serverless databases overview
- 23:20Cost challenges in serverless databases
- 28:25Building a serverless database
- 31:20Serverless cluster orchestration
- 33:30Client connection and recovery
- 35:58Cold starts in serverless databases
- 37:27Handling server shutdowns
- 39:41Final takeaways and Q&A
related talks
01h 00mServerless - reality or BS - notes from the trenches
47mAdvanced Serverless Architectural Patterns on AWS
30mPerformance Tales of Serverless
25mWhere Did All My Servers Go?
35mEfficiency of Serverless Architectures with AWS Solutions
28mLegendary Lambdas
58mAWS Lambda and Serverless Framework Deep dive
53mDissecting Serverless Enterprise Java Applications #slideless