Comments on: Benchmark of Python WSGI Servers

By: Rob

Rob — Sat, 21 Jan 2012 14:27:36 +0000

Bjoern is really fast.. i’ve done a simple test with bottle app (bottlepy.org) with only one view that returns “hello world”.

I used apache-bench as client, with 100 concurrent connections and without keep-alive (ab -c 100 -n 10000 http://…..).

Results:

bjoern: Requests per second: 5153.34 [#/sec] (mean)
fapws: Requests per second: 3135.04 [#/sec] (mean)
cherrypy: Requests per second: 1795.33 [#/sec] (mean)
tornado: Requests per second: 1752.45 [#/sec] (mean)
gunicorn: Requests per second: 1696.98 [#/sec] (mean)
diesel: Requests per second: 1540.02 [#/sec] (mean)
gevent: Requests per second: 1488.87 [#/sec] (mean)
rocket: Requests per second: 1169.62 [#/sec] (mean)
twisted: Requests per second: 611.43 [#/sec] (mean)

By: Riccy Yu

Riccy Yu — Thu, 19 Jan 2012 03:31:07 +0000

Great job, excellent reference for easily choosing a web server framework!
At now, i use nginx+fcgi for my web server, but i think i should have a try for uwsgi or gevent!
It is really a great article for all!

By: James

James — Tue, 10 Jan 2012 20:14:06 +0000

Thanks for the thorough benches and write-up. I know this was a while ago, but can you provide any more details on the setup of your virtualized environment? i.e. were you using paravirtualization (probably if using Xen), or Hardware assisted (e.g. Intel VT-x on VMWare, KVM etc.). This will have a pretty large effect at the top-end; possibly some of the servers were limited more by the network throughput/PPS limit than their own capabilities.

By: webguo在路上 » 高性能python web服务器

webguo在路上 » 高性能python web服务器 — Fri, 06 Jan 2012 12:41:58 +0000

[...] app服务器决定着整个系统的响应快慢。通过参考Nicholas Piel写的《Benchmark of Python WSGI Servers》我圈定了以下几个服务器（模块）：mod_wsgi for [...]

By: Quora

Quora — Fri, 23 Dec 2011 10:17:20 +0000

Why did TellApart write TAFE instead of just using Tornado or any of the existing pure Python webservers?…

It should be cleared up that TAFE stands for TellApart Frontend End and, as described in Mark Ayzenshtat’s blog post (mentioned in the question,) is a framework built on the python coroutine library gevent. The decision In general to move from apache/…

By: overtip

overtip — Fri, 23 Dec 2011 06:54:38 +0000

mmmmmm

By: ksamuel

ksamuel — Wed, 19 Oct 2011 12:26:16 +0000

Interesting benchmark. I would use uwsgi if it wasn’t such a pain to configure.

I would like to point out that we do server a 120 000 users a day using gunicorn but:

- we use several workers (around 12). Gunicorn is not meant to be used with one worker.
- we use nginx in front.
- above all, we use the “–worker-class=”egg:meinheld#gunicorn_worker” options that set gunicorn to use a different worker class, much, much faster.

By: Anonymous

Anonymous — Thu, 25 Aug 2011 10:22:30 +0000

Do you want to test out the same with the latest versions of the servers that you used here? It’s more than 1 yr since your results and an update will be great. That will also show how active the respective developers are.

By: Quora

Quora — Tue, 23 Aug 2011 04:44:09 +0000

What are best-practices for deploying a web app with PyPy? (gunicorn, tornado, etc?)…

After reading a very interesting thread on Python Stacks in Hacker News (http://news.ycombinator.com/item?id=2910953), I decided to give uwsgi a try and it seems to perform incredibly well, which I can vouch myself as far as my limited testing since ye…

By: 高性能python web服务器

高性能python web服务器 — Wed, 17 Aug 2011 08:40:28 +0000

[...] app服务器决定着整个系统的响应快慢。通过参考Nicholas Piel写的《Benchmark of Python WSGI Servers》我圈定了以下几个服务器（模块）：mod_wsgi for [...]