LOL how in the world js nanoexpress have better performance than Go or Go Fiber

[kamalshkeir]

For helloworld project, i did the test on my machine and everytime i get the same result, better performance for GoFiber.
GoFiber handle 77k req/s
nanoexpress handle 50k req/s

[waghanza]

Hello @kamalshkeir,

Thanks for concern.

First of all, I will say that this project is still in beta stage.

I understand that you are concern about fiber vs nanoexpress. I'll try to explain the differences in 2 points :

• as you noticed https://web-frameworks-benchmark.netlify.app/result?f=nanoexpress,fiber are not corresponding to any above figures -> the idea is to have a comparison in an architecture corresponding to what IS server (e.g : production server / vms / pods)
• we are using docker -> I'm not sure we can correlate any real-word figure with local docker
• are we sure we use the same version / go compilation options / os ?

Regards,

[kamalshkeir]

I was shoked to see a dynamic language like Js beat a compiled language like Go, i have test my Go server against .Net , Django, Ruby..., never seen any server that can handle 70k req/s like Go do.
Anw i understand how hard to simulate a real world benchmark that work the same on every machine.
btw i was using go 1.16.6 on windows machine localy vs nanoexpress on the same machine locally without docker for my benchmarks.
Thank u for ur answer

[waghanza]

Very hard explain those results differences, but what I can say is :

• networks are NOT handled the same way on unix and windows (I'm not sure their is any value to explain those gaps)
• a interpreted language could be faster for a particular benchmark -> for example, most of stack use advanced techniques (jit, low lever wrappers ....) that infirm this assertion

[re-thc]

The test isn't really testing javascript vs go. It only tests the web server component.

It makes sense that nanoexpress has comparable performance to fiber. Besides a very minor part of the code, nanoexpress uses heavily optimized c/c++ in the form of uWebSockets, which would be faster than Go.

If you're interested in making a point you can try Deno (based also on javascript / v8) with its new Rust based Http server and that might be even faster :).

[waghanza]

The test isn't really testing javascript vs go. It only tests the web server component.

indeed, comparing languages could only be made using the same kind of component, here (server-side) frameworks.

It makes sense that nanoexpress has comparable performance to fiber. Besides a very minor part of the code, nanoexpress uses heavily optimized c/c++ in the form of uWebSockets, which would be faster than Go.

Right, some frameworks uses some low level components (mostly to deal with network) such as https://www.swoole.co.uk in php, https://github.com/jonashaag/bjoern in python, https://deno.land/ in javascript ...

The point is that any language COULD (in theory) be faster than expected. It depends mainly on code design (async/sync, complexity, abstraction ...)

[waghanza]

Please not however, that the actual results are still in beta. I mean no real network is used (only loopback interface) and then results probably would vary a lot after running on a real-life network

[bangbaew]

I've tried load testing using K6 (64 VUs, 10s), from left to right: Nest.js Fastify, Gofiber, Nanoexpress
and Nanoexpress gave me impressive results.
The syntax is almost identical to Express, just change from app.delete method to app.del,
I've simply changed only require('express') to require('nanoexpress') and it runs normally.

For the second, third and fourth try, Nanoexpress had a little drop in performance but still beat Gofiber

[sharpninja]

Are the official test result run in GitHub Actions?

[waghanza]

nop. their are running on a private workstation. the choice was made since this workstation has closer specs facing some cloud servers (the idea behind that is to run a a real cloud when this project will be mature and in production)