Serverless 2049
summary
Performance tests run an event pipeline on AWS Lambda comparing Plain Java (JVM) and GraalVM native images with Micronaut handlers, splitting large inputs (up to 500 MB files) into 50 MB chunks. GraalVM shows faster cold starts but a surprising S3 download slow-down that makes it slower on some end-to-end runs (10 files: Java 260 s vs Graal 180 s) and, when extrapolated, can be more expensive overall; Lambdas often account for only ~10–15% of monthly serverless costs.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:59Why serverless 2049
- 8:11Serverless cost and quality
- 16:18Benchmarking languages
- 24:04Total cost of ownership
- 28:20Data pipeline architecture
- 33:00Lambda Power Tuning
- 41:54Java vs GraalVM in production
- 52:12Research findings and plans
- 54:38Q&A on SnapStart
- 1:00:40Future enhancements
related talks
53mScale up with GraalVM and AWS Lambdas
53mHigh performance Serverless Java on AWS
24mHigh Performance Serverless Functions in Scala
46mThawing Java on AWS Lambda Reducing Cold Start Times from 11 Seconds to 1
51mAdapting Java for the Serverless world!
56mAfraid of Java cold starts in Serverless? Fear not, Java is fast
34mFast and Efficient Java Microservices with GraalVM
52mMaximizing Performance With GraalVM