Gevent compared to Erlang, Java (Webbit),Go, Node.js

[Allard]

I haven't delved in to the whys an interesting observation:

https://github.com/ericmoritz/wsdemo/blob/master/results.md

The benchmark tested an echo like scenario and compared Erlang, Java (Webbit),Go, Node.js and Python (gevent + ws4py). The Python version of the test seems to be the least favourable.

It may get some comments on Hacker News as well.

http://news.ycombinator.com/item?id=4105317

Cheers,

Allard

[Executier Godlike]

> I haven't delved in to the whys an interesting observation:
> https://github.com/ericmoritz/wsdemo/blob/master/results.md

Could you add plots for easier comparison?

> The Python version
> of the test seems to be the least favourable.

Sure, how python can compete with compilers and JIT-enabled
staticaly-typed languages? :)

--
Alexandre Kandalintsev

[Benoit Chesneau]

erlang isn't one of those ^^

[Antonin AMAND]

Personally I'm already thrilled that it stays in the same order of magnitude.
I would like to see how ruby-eventmachine compares to gevent.
Does anybody knows some benchmarks comparing them ? (not necessarily
on websockets)

[Matt Billenstein]

Go, Java, Erlang are probably not limited to a single core like node/python?

gunicorn with many gevent workers would fare much better I think - and be a
more realistic deployment option for anyone serious about these sort of
workloads under Python.

And there are options [1] in C++ using fibers which would be interesting to see
- I'd imagine there is a possibility some of these could be much faster than
Erlang.

m

1. http://code.mozy.com/projects/mordor/

On Wed, Jun 13, 2012 at 04:49:53AM -0700, Allard wrote:
> I haven't delved in to the whys an interesting observation:
> https://github.com/ericmoritz/wsdemo/blob/master/results.md

> The benchmark tested an echo like scenario and compared Erlang, Java

> (Webbit),Go, Node.js and Python (gevent + ws4py). The Python version of

> the test seems to be the least favourable.

> It may get some comments on Hacker News as well.
> http://news.ycombinator.com/item?id=4105317
> Cheers,
> Allard

--
Matt Billenstein
ma...@vazor.com
http://www.vazor.com/

[Jeff Lindsay]

I feel very guilty as the original author of the ws4py's gevent modules. I never got around to optimizing or really stabilizing it. :(

--
Jeff Lindsay
http://progrium.com

[Allard]

Agree that it would have been good to see ruby's eventmachine in here and a C++ example. Using multiple workers with Gunicorn might have been more fair too. I am wondering what the hot spots are in the code. If it's even ws4py or if it's somewhere else...

On Wednesday, 13 June 2012 14:59:55 UTC-4, Jeff Lindsay wrote:

I feel very guilty as the original author of the ws4py's gevent modules. I never got around to optimizing or really stabilizing it. :(