Tackling Dreaded .Net Cold Starts When Going Serverless
summary
Claim: .NET shows the worst cold starts on AWS Lambda but targeted fixes cut them dramatically. Experiments show memory tuning lowers cold-start time while raising teardown frequency, scheduled pings reduce cold-start rate from ~19.41% to ~0.77% with five‑minute triggers, and moving dependencies into lambda layers plus using the dotnet CLI to trim assemblies improves startup. A custom runtime using CoreRT (lambda native) yields roughly 10× faster cold starts in tests, at the cost of harder local development.
AI-written summary · about
chapters (8)
- 0:00Start
- 1:22Agenda
- 9:34Motivations
- 15:02INHERENT PROBLEM OF SERVERLESS
- 21:58STATICALLY TYPED LANGUAGES
- 26:11LIFECYCLE OF A BERVERLESS FUNCTION
- 35:05UBAGE OF LAMBDA LAYERS
- 37:01KEEPING CONTAINER WARM
related talks
51mHow To Reduce Cold Starts for Java Serverless Applications in AWS
19mTen things that improve the performance of your Java based AWS Lambda
27mServerless Performance on a Budget
35mEfficiency of Serverless Architectures with AWS Solutions
46mReduce Cold Starts for Java Serverless Applications in AWS
32mServerless & JVM: Build Your Benchmark to Bust Preconceptions
50mAWS Serverless with .NET Core
44mHow to develop, run and optimize Spring Boot 3 application on AWS Lambda