httpd-test-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Ames <grega...@apache.org>
Subject Re: mod_specweb99 debugging...
Date Fri, 16 Aug 2002 14:01:04 GMT
Chris Johnson wrote:
>
> Hi:
> I'm using your specweb99 modules to examine apache performance, and was
> wondering if you have newer versions handy - I noticed that the 1.3
> version did not compile on rh7.2
> (the SEMCNTR define needed to be disabled); it seemed to work
> afterwards. 

Would you mind sending a patch to test-dev@httpd.apache.org?  We like diff -u. 

You might want to subscribe to test-dev@httpd.apache.org (developer discussions)
and test-cvs@httpd.apache.org (automatic cvs commit notifications).  There are a
couple of other subprojects supported there, but they are both very low volume
mailing lists.

> The 2.0 version has problems with dynamic content (lots of invalid and expired
> ads).

I fixed that one.  You see those messages on the very first pass after the
client tests the various types of requests, and then they go away on subsequent
passes, right?  If so, you can just do a "cvs up" on httpd-test to get that
fix.  I assume you know how to use anonymous cvs?

> Being new to apache debugging, any info on debugging these modules would
> be greatly appreciated.
> 
> thanks
> 
> P.S. I'm hoping to analyze kernel bottlenecks (using lockmeter, cpu
> statistics, etc.) after examining optimizing httpd in user space.

Sounds great!  I hope you have a *lot* of client horsepower and bandwidth, or
else a very wimpy server with a lot of disk that can run current software. 
Otherwise it's hard to drive the server to saturation with this workload.

Naturally, we'd like to see any interesting user space results you come up with
on dev@httpd.apache.org (especially for 2.0) or dev@apr.apache.org, as
appropriate.  Some Linux kernel hackers I work with at IBM are also doing
specweb99 benchmarks with 2.0.  They are seeing returnHTMLPageWithBuffer (a
mod_specweb99 function) as the hottest user space function.  But they also have
issues with oprofile overflowing some hash table, so I don't know whether to
believe it or not.  No smoking guns jump out at me when I inspect that code, so
independent confirmation would be nice.

Greg

Mime
View raw message