httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1)" <mad...@hp.com>
Subject RE: Regarding worker MPM and queue_push/pop
Date Sun, 07 Dec 2003 16:43:18 GMT
Hi,
	Just wanted to update y'all - I traced down the bottleneck to
mod_specweb99.c. The bottleneck is caused by the POST & CAD_GET
transactions. If I eliminate the POST & CAD_GET from the SPECweb99 requests,
I don't see any spiky activity - the CPU usage is steady at 100%.

	More comments below ..

> -----Original Message-----
> From: Aaron Bannert [mailto:aaron@clove.org]
[SNIP]
> 
> I can't imagine why sleeping would consume 20% of your CPU when the
> server is idle. Perhaps you should look into the scalability of your
> thread library. On what kind of system are you seeing this problem
> (hardware/OS/rev/etc)?

I'm using MxN threads on HP-UX 11.23 - running on a IA64 madison chip
(1-way). I think the problem was because my build was using semop's in some
places even if pthread_mutex were available. AIUI, pthread_mutex performance
(for process-wide locks) is better than semop performance.

-Madhu

Mime
View raw message