Why You Don't Need to Worry about Scaling Your Java Webapp
Talkat JCON Europe 2023 · on Thursday, 2023-06-22· at 11:00· Cinema Blackbox, Im Mediapark 1, 50670 Köln, Germany·
summary
A Spring Boot 3 app on Jetty with one JSON endpoint is hit by a custom Java load tool using 4 loaders + isolated probe and an HDR histogram to measure latencies at 1k, 5k, 7.5k and 20k req/s. OS metrics and async-profiler flame graphs show the controller costs at ~0.13% CPU, Jackson ~6%, heavy JIT compilation, and in a Spring PetClinic run Thymeleaf consumes ~60% CPU causing 2–8s latencies—so the process finds the real bottleneck and what to scale.
AI-written summary · about
chapters (10)
- 0:00Introduction to scaling
- 2:08Load testing methodology
- 5:52Running the load test
- 8:56Data collection and analysis
- 14:51Interpreting test results
- 20:22Using flame graphs
- 26:21Case study: Pet Clinic
- 32:14Analyzing scaling limits
- 34:33Practical advice and outlook
- 43:33Q&A session
related talks
39mHow to Reduce the Footprint of Your Spring Boot Applications
52mLean Spring Boot Applications for the Cloud
48mLean Spring Boot Applications for the Cloud
01h 03mCloud Cost Optimization for Java Developers
54mHow Fast is Spring
51mLean Spring Boot Applications for The Cloud
49mBuilding Lean Spring Boot Applications for the Cloud
38mPractical Performance Analysis