Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Improve techempower benchmark

See original GitHub issue

So, Javalin is back here: https://www.techempower.com/benchmarks/#section=data-r21&l=zik0vz-6bj&test=composite

But it’s not looking so fast. Maybe we can fix that?

The repo for the benchmark code is here: https://github.com/TechEmpower/FrameworkBenchmarks/tree/master/frameworks/Java/javalin

Some things we should do:

Update to newest Javalin
Replace Pebble with JTE and precompiled templates
Remove before/after/error stages from request lifecycle
Cleanup database controller
Use custom JSON setup
Enable Loom via --enable-preview and Java 19
Copy DB stuff from vertx or jooby?
~Enable HTTP2~ (not supported by wrk)
~See if we can make Jackson faster~

Here is our branch: https://github.com/javalin/FrameworkBenchmarks/tree/javalin-improvements

Issue Analytics

State:
Created a year ago
Comments:26 (26 by maintainers)

Top GitHub Comments

1reaction

zugazagoitiacommented, Oct 8, 2022

@zugazagoitia is there a warmup phase before the benchmark? If so, the JIT should probably beat AOT GraalVM compiler in terms of throughput.

There is a warmup, I have no idea if this would improve performance but either way I’ll make a graalvm branch and run some benchmarks, you might be right.

looked into vertx-web and jooby, and they both have -XX:+AggressiveOpts JVM option set. Maybe we could give it a try how this affects the benchmark?

After a litte googling I found this: https://bugs.openjdk.org/browse/JDK-8199847

Looking at the results, it seems the DB is also a bottleneck at the moment. I tried to look into how other frameworks connect to the DB. The frameworks with high scores I looked at (officefloor, inverno, jooby) all use vert.x for DB connection handling. Maybe it would be possible to replace Hikari with https://github.com/eclipse-vertx/vertx-sql-client?

I believe the DB to be the biggest bottleneck right now, we have major improvements comming to Json on a dsl-json branch I’m working on. Could you try and replace it?

Also, I’ve enabled actions for the repo. They (TechEmpower) encourage forks to use the workflow to test benchmarks, it can verify that the endpoints provide valid output.

1reaction

tipsycommented, Oct 8, 2022

@zugazagoitia a lot of people would appreciate that in general, the tutorial is outdated.

Top Results From Across the Web

Round 21 results - TechEmpower Framework Benchmarks

Rnk Framework Best performance (higher is better) Best performance (higher is better) Er... 1 drogon‑core 616,607 100.0% 0 2 xitca‑web 587,955 95.4% 0 3 drogon 556,046...

JSON performance in TechEmpower benchmarks · Issue #1177

The latest TechEmpower benchmarks have been released. Axum is doing extremely well and is in the top ten.

TechEmpower Framework Benchmarks - Twitter

Continuous Benchmarking results for the TechEmpower Web Framework Benchmarks Project. ... a 70% speed increase on TechEmpower Fortunes benchmarks ...

TechEmpower benchmarks for opium running httpaf are ready

Source for the TechEmpower Framework Benchmarks project - TechEmpower/ ... Am I doing something wrong, how can I get better performance?

I was looking at TechEmpower benchmarks, and noted that ...

34 votes, 55 comments. Hello, maybe you can explain why PHP (php-ngx) seems to be better than Node. Am I reading the score...