Hey Luke, first of all awesome project.
Its not an issue but a comparison with a complete different framework of a different language. Since I want to cut down server cost & all & provide an API server for a SAAS but I work with Node JS & have only little experience with Python.
So I was looking at the project Japronto its freakishly fast. More info about it here. I've checked the benchmarks & I don't understand which ones to compare between Japronto & Polka.
Polka benchmarks says 41053.98 & then 414739 requests in 10.10s, which ones to compare ❓
Can Polka be as fast as Japronto like it says it gives 1 million requests/second❓
Hey, thanks!
Japronto is mostly written in C, it just has python bindings. So no, Polka can't be as fast as a C-based library.
If you're weighing between languages (not just frameworks), have a look at which_is_the_fastest. It's a pretty good suite.
Otherwise, if you're sticking with node.js, Polka is the fastest of the options, as it's within 2% of a native & naked Node.js server.
Hope that helps!
Thanks @lukeed, didn't knew about that link
You bet! Was in your shoes too, before deciding to build Polka lol
Curiously, Polka is faster than Japronto (almost twice as fast), according to which_is_the_fastest.
Another point is that Polka will shine even more if it is possible to use it in cluster mode, since the node part of the benchmark is running in a single thread.
I heavily suspect that's a bug with the new benchmark suite 😆That repo has undergone a bunch of changes in the last week or so, and Japronto got caught in the mix.
But yes, +1 re:cluster mode. On my machine (and Node 8.9), Polka goes from 42k req/s to ~95k req/s with simple benchmarks.
@lukeed @paulocoghi the benchmark suite above is actually unrevelant (I know it, I maintain it). Soon, results will be more accurate, I'm working on it => https://github.com/the-benchmarker/web-frameworks/pull/632 :heart:
And node are running with pm2 => not on a single thread then
It's the only way I've found to makenode implementations uniform 😁
Any ideas are welcome :heart:
Most helpful comment
I heavily suspect that's a bug with the new benchmark suite 😆That repo has undergone a bunch of changes in the last week or so, and Japronto got caught in the mix.
But yes, +1 re:cluster mode. On my machine (and Node 8.9), Polka goes from 42k req/s to ~95k req/s with simple benchmarks.