Performance Problems.. Server hardware smoked by $500 box?

[Adam C. Migus]

Robert Watson wrote:

>On Thu, 11 Sep 2003, John Straiton wrote:
>
>  
>
>>(Production WEB now that it's using apache2. Since we're new to apache2,
>>it's gonna be certainly hokie)
>>    
>>
>
>My understanding is that Apache2 is optimized to make extensive use of
>threads.  On FreeBSD 5.x, the pthreads implementation (libkse) is
>substantially better than in 4.x, especially in SMP environments.  In
>FreeBSD 4.x, I'm generally suggesting to people that they use Apach1
>unless there are specific feature requirements that make Apache2 the only
>choice.  On FreeBSD 5.x, I haven't really evaluated Apache2 behavior, but
>it could well come down to the threading issue, and may well be a lot
>faster than Apache1 when linked against the right threading library.
>
>This is all speculation of course, but it could be that differences in
>Apache version make up most of the performance difference you see.
>
>Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
>robert at fledge.watson.org      Network Associates Laboratories
>
>_______________________________________________
>freebsd-performance at freebsd.org mailing list
>http://lists.freebsd.org/mailman/listinfo/freebsd-performance
>To unsubscribe, send any mail to "freebsd-performance-unsubscribe at freebsd.org"
>  
>
Actually, if John is using the apache2 port, without specifying a custom 
MPM, he's using the prefork MPM.  The apache2 prefork MPM is 
non-threaded and uses a method similar to apache1 to serve requests.

Thus threading is irrelavent unless he's specifically enabled them, 
which, he cannot be using since he's using PHP, which doesn't work with 
apache2 threading.  I use apache2 on FreeBSD 5.1 RELEASE and find 
performance to be as good or better than apache1.

I suspect the excessive numbers in the prefork section might be causing 
some trouble and would recommend reverting back to the defaults (with 
the exception of MaxClients) for your apache2 configuration (see the 
documentation for details).  Also try using a local, caching bind setup 
for name service if you find performance is still lacking.  For the nfs, 
I'd also mount with rw,rdirplus.

Adam