Hacker News Reader

Comment by derkoe | Hacker News Reader

Our Spring Boot showcase application https://github.com/porscheinformatik/angular-spring-heroes starts ~ 10% faster on Java 17 compared to Java 11 (compiled with target 11).

View on HN

Even more interesting is the memory usage (app started and opened index page in browser):

  - Java 8/11 uses ~ 550MiB
  - Java 17 uses ~ 260MiB

(all run with Docker and a memory limit of 1GiB)

What GC is used, and how much CPU time/quota is given to the container?

Default GC, and no quota - which means my full machine (i7-6700HQ)

Just tried a simple Quarkus app (https://github.com/derkoe/quarkus-htmx-todos) with the same results. Java 17 starts around 10% faster. Memory consumption is around 10% lower (after first request).

I hope spring moves to compile time di like quarkus and ajave inject instead of scanning the classpath at runtime.

You mean Spring native [1], i think it's not ready yet.

[1] https://docs.spring.io/spring-native/docs/current/reference/...

No I do not. I do not want native image. I want the dependency injection container created at compile time like Dagger2 [0] or Avaje Inject [1] or quarkus without native image.

[0] https://dagger.dev [1] https://avaje.io/inject/

You can use the Spring AOT plugin [1] without enabling the support of GraalVM native image

[1] https://docs.spring.io/spring-native/docs/current/reference/...

Ohhh I see. Thank you!

Isn't that what Micronaut does, from the creators of Spring? I never used the latter, but I really like working with the former.

Yes, Micronaut and Quarkus do that. Micronaut was created by a company (https://objectcomputing.com/)

Do you happen to have a comparison from Java 8 to 11?

Just tried it - there is no significant difference between Java 8 and 11! They both start around 6.5s (JVM running for 7.4s). Java 17 starts the app in approximately 5.9s (JVM running for 6.7s).

We did java 8 vs java 11 a few years ago: https://www.optaplanner.org/blog/2019/01/17/HowMuchFasterIsJ...

TLDR: using the same GC (*), it's 4% to 16% faster (depending on the GC).

But be warned that if you don't set a GC, Java 8 uses Parallel GC (high throughput, so faster) by default and Java 11 uses G1 GC (low latency, so less hickups) by default.

Reply on news.ycombinator.com