Here are some preview results from August of TechEmpower benchmarks, Java or Kotlin, Micro or Full f...

# server

[orangy]

Here are some preview results from August of TechEmpower benchmarks, Java or Kotlin, Micro or Full frameworks (excluding raw platforms). A little bit more work and I think Kotlin frameworks can climb higher 🙂
https://www.techempower.com/benchmarks/previews/round15/#section=data-r15&hw=ph&test=json&l=xan9tr&c=4

[dave]

We've actually been having real trouble with the TFB. There's something very weird going with the http4k results - for both preview rounds so far the apps have basically refused to start and provide any results - something which I can't reproduce locally. Initially I thought it was just round 1 but then it happened again with R2. For the moment I've stripped out the database routes to see if that makes a difference for R3. It also doesn't really help that the rounds are so far apart.. 😕

[orangy]

Yeah, I totally agree. Iterations are non-deterministic and almost no diagnostics. We have hard time with them also…

[orangy]

Anyway, http4k-jetty is faster than #spring :))

[napperley]

Good to see Vert.x in the top tier (always excels at JSON parsing/serialisation, it is one of its specialities), now it would be good to see the Kotlin ver in the benchmarks 😉. I see that Play is in the bottom tier. Lets pretend that Scala is being used with Play instead of Java 🧌. Most people when they use Play go for the Scala ver anyway.

[napperley]

Grizzly (basic web server that was originally part of Glassfish) like Vert.x in the benchmarks is a consistent top performer, except when Grizzly is used with Jersey which seems to bring performance down a lot due to all that Jersey magic ✨.

[julienviet]

Techempower benchmark are really difficult to apprehend

[julienviet]

we see results that don't make sense, i.e we provide vertx-web and vertx and in some case vertx-web has the same perf than just vertx

[julienviet]

and we know that vertx perf should be much better than vertx-web

[julienviet]

also most benchmark don't benefit to non-blocking servers