Frameworkbenchmarks: Python: tornado with asyncio(uvloop(libuv))

[westurner]

Here's a probably pretty fast stack: Tornado with asyncio with uvloop (libuv; (Node, ))

• tornado with asyncio:
  
  
  
  • src: https://github.com/tornadoweb/tornado
  
  
  • src: https://github.com/tornadoweb/tornado/blob/master/tornado/ioloop.py (epoll, kqueue, select)
  
  
  • src: https://github.com/tornadoweb/tornado/blob/master/tornado/platform/asyncio.py (asyncio(...))
  
  
  • docs: http://www.tornadoweb.org/en/stable/asyncio.html
  
  
  • docs: http://www.tornadoweb.org/en/stable/wsgi.html#tornado.wsgi.WSGIAdapter
  
  
• asyncio:
  
  
  
  • src: https://github.com/python/cpython/tree/master/Lib/asyncio
  
  
  • src: https://hg.python.org/cpython/file/tip/Lib/asyncio
  
  
  • docs: https://docs.python.org/3/library/asyncio.html
  
  
• uvloop
  
  
  
  • src: https://github.com/MagicStack/uvloop
  
  
  • docs: https://uvloop.readthedocs.io/
  
  
• libuv
  
  
  
  • src: https://github.com/libuv/libuv
  
  
  • docs: http://docs.libuv.org/en/v1.x/
  
  

[westurner]

See: #1567 (REQ: Gunicorn w/ greenlets)

[westurner]

uvloop w/ httptools looks pretty fast (compared with nodejs and golang, for example) in these charts:
https://magic.io/blog/uvloop-blazing-fast-python-networking/

...

• sanic

  • src: https://github.com/channelcat/sanic/
  • pypi: https://pypi.python.org/pypi/Sanic
  • docs: https://sanic.readthedocs.io/en/latest/
  • sanic is built on asyncio, uvloop (libuv (node,)), and httptools (node)

• httptools

  • src: https://github.com/MagicStack/httptools
  • src: https://github.com/MagicStack/httptools/blob/master/tests/test_parser.py
  • pypi: https://pypi.python.org/pypi/httptools
  • desc: "httptools is a Python binding for nodejs HTTP parser. [...]"
  • httptools.HttpRequestParser, httptools.HttpResponseParser, httptools.parse_url

[westurner]

Japronto and Meinheld look pretty fast for this "Hello World" benchmark:

https://medium.freecodecamp.com/million-requests-per-second-with-python-95c137af319

Most of the servers execute requests from pipelining clients in the same fashion they would from non-pipelining clients. They don’t try to optimize it. (In fact Sanic and Meinheld will also silently drop requests from pipelining clients, which is a violation of HTTP 1.1 protocol.)

.

• japronto

  • src: https://github.com/squeaky-pl/japronto
  • docs: https://github.com/squeaky-pl/japronto/blob/master/README.md
  • uvloop, picohttpparser

• picohttpparser

  • src: https://github.com/h2o/picohttpparser
  • src: https://github.com/h2o/picohttpparser/blob/master/picohttpparser.c
  • docs: https://github.com/h2o/picohttpparser/blob/master/README.md

[westurner]

See: https://github.com/TechEmpower/FrameworkBenchmarks/pull/2311#issuecomment-259327909 (re: Python performance)

[cjnething]

Hi @westurner,
We are not experts on every language/framework, so we rely on the community to make updates and add new frameworks. If you see areas of improvement, we encourage you to make a pull request! Thanks!

[nbrady-techempower]

what @cjnething said. Thanks for the well thought out issue/resources. Would love to see how some of these implementations would fair in the benchmarks.

[westurner]

https://morepypy.blogspot.com/2017/03/async-http-benchmarks-on-pypy3.html

[GMLudo]

Hello people,

I'm working to integrate all of these new tools in FrameworkBenchmarks.
Too late for round 14, it should be good for round 15.

Have a nice week-end.

[westurner]

:+1:

On Saturday, March 25, 2017, Ludovic Gasc [e-mail] wrote:

Hello people,

I'm working to integrate all of these new tools in FrameworkBenchmarks.
Too late for round 14, it should be good for round 15.

Have a nice week-end.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/TechEmpower/FrameworkBenchmarks/issues/2560#issuecomment-289215680,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AADGy0ChCndXGNSohzWDvORuhLRNIc8Lks5rpSbFgaJpZM4L7xkY
.