Language Runtimes for the New Cloud Era
summary
High‑level language runtimes impose 10–100 MB footprints and 100–200 ms cold start penalties while many serverless functions run under one second and are billed per millisecond, so startup and memory overhead dominate cost. Measurements show loading an 854 MB text can use ~6 GB RAM; KVM/Firecracker and containers cost ~100 MB/100 ms. A prototype runtime-sandbox that consolidates many functions onto one VM reduces redundant initialization and cold starts but exposes hard corner cases: snapshot timing, JIT warmup/deoptimizations, GC transparency and forking/thread revival.
AI-written summary · about
chapters (8)
- 0:00Language runtimes in the cloud
- 2:12High-level languages performance
- 6:48Serverless runtime challenges
- 10:01Optimizing runtimes and cold starts
- 18:06Snapshotting inefficiencies in serverless
- 23:07Memory footprint overhead
- 26:43Future directions for runtimes
- 28:30Q&A runtime sharing
related talks
51mAdapting Java for the Serverless world!
32mServerless & JVM: Build Your Benchmark to Bust Preconceptions
42mServerless Systems That Won't Burn a Hole in Your Pocket
46mMonolith to Serverless - Can Modern Cloud-Native Java Runtimes Do It All
53mScale up with GraalVM and AWS Lambdas
46mReduce Cold Starts for Java Serverless Applications in AWS
27mServerless Performance on a Budget
54mContainerless in the cloud with AWS Lambda