On Wed, Oct 3, 2012 at 10:35 AM, Niphlod <niph...@gmail.com> wrote: > Well, seems to work. I'd need to know if there is some way to test if all > web2py framework works on on this, but a normal app seems to do pretty fine. > Windows Vista (aaargh!), PortablePython 2.7.3 32-bit, T8100 2.10Ghz, 3 GB > RAM, adjusted softcron in anyserver.py (to allow removal) in order to keep > things smooth, removed logging of every request in the console of tornado > (commented out line 2214 of motor.py) > > Performance-wise, same app started with anyserver, one loads cherrypy and > the other tornado, 2 concurrent benches (one for static files, the other for > the "hello world" app) no memory leaks for either. > > ab -c 100 -n 1000 (it is windows after all, bumping to -c 1000 -n 1000000 at > this point is unuseful) > > /app/default/index rps cherrypy 14.90 , tornado 132 > /app/static/test.css rps cherrypy 23.71, tornado 174
I guess cherrypy spawns some threads, right? To be fair can you ajust to use only one thread/process to match tornado. I think you will see some improvements in cherrypy rps in this particular bench, since the OS/Python threads switching contexts is avoided. I think web2py should not choose a webserver only in terms of speed, but should focus on the simplest, complete and pure python webserver, My 2 cents. Ricardo --
