[mod_python] need help solving some performance issues

>From my experience, with worker MPM not only multiple cores helps, but I
think larger L2 cache helps a lot.
I suspect it has to do with context switching- perhaps data from several
threads can be kept in L2.

I have two almost identical servers serving same requests (load balancing),
first with 2x 2.66GHz Xeons with 512KB L2 , and second with 2x 3.0GHz Xeons
with 1MB L2. The performance of the second server is much better than what I
would expect from 2.66/3.0 GHz ratio.

> -----Original Message-----
> From: mod_python-bounces at modpython.org [mailto:mod_python-
> bounces at modpython.org] On Behalf Of Graham Dumpleton
> Sent: Thursday, April 10, 2008 3:48 PM
> To: Culley Harrelson
> Cc: mod_python at modpython.org
> Subject: Re: [mod_python] need help solving some performance issues
> 
> 2008/4/11 Culley Harrelson <harrelson at gmail.com>:
> >
> >  I have no experience with dual cpu machines though-- this is part of
> my
> >  concern.  I manage 6 production FreeBSD servers but they are all
> single cpu
> >  systems.  If apache is in prefork mode 2 cpus is just extra
> processing power
> >  right?!?
> 
> Even if you were using worker MPM and multithreaded applications,
> multiple CPUs or cores will help with Apache, because it would still
> be running as a multiprocess application and a lot of stuff happens in
> C code also which is capable of being spread across multiple CPUs or
> cores.
> 
> For Apache on UNIX boxes at least, the Python GIL is not the big
> problem that the many people make it out to be. See:
> 
>   http://blog.dscpl.com.au/2007/09/parallel-python-discussion-and-
> modwsgi.html
>   http://blog.dscpl.com.au/2007/07/web-hosting-landscape-and-
> modwsgi.html
> 
> Graham
> _______________________________________________
> Mod_python mailing list
> Mod_python at modpython.org
> http://mailman.modpython.org/mailman/listinfo/mod_python