Honey, I stress tested shrank images!
summary
Compares a single travel-service app implemented as Spring MVC, Spring WebFlux and Quarkus, with Postgres prefilled (≈500k transfers, 200k users), Gatling-style load, and images built via JIB, buildpacks and GraalVM native. Results show native builds take 40–60× longer but produce instantish boots (~100 ms) and much lower RSS, WebFlux gives ~2× RPS versus servlet, non-native Docker images sit near ~300 MB while Quarkus native distroless ≈70 MB (manual microbase ≈17 MB), and Docker seccomp causes measurable slowdown unless tuned.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 5:09Talk begins
- 6:13Agenda
- 7:19Research questions
- 10:54Database setup
- 13:51API-first approach
- 17:25Spring Servlet results
- 19:58Spring Native results
- 21:27Spring Reactive results
- 23:34Quarkus results
- 27:56Quarkus vs Spring Native
- 31:53Docker tuning
- 33:18Summary
- 34:44Recommendations
- 36:09Q&A
- 36:50Developer experience
- 38:38Native images
- 41:29Framework market
- 42:33Closing
related talks
59mQuarkus for Spring Developers
46mMigrating a real‑world application to Quarkus
30mSupersonic Java Quarkus with Knative serverless
49mQuarkus Intro
52mQuarkus Supersonic Subatomic Java
54mQuarkus: Java development turned into delight
51mKubernetes Native Spring apps on Quarkus
48mSupersonic, Subatomic Java with Quarkus