httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Robert Jones" <robe...@cs.pdx.edu>
Subject Apache2 as a Layer7 webswitch - performance problems compared to 1.3?
Date Sat, 18 May 2002 19:25:16 GMT
I have set up an Apache 1.3 content-aware web switch for a Linux/Apache
cluster using the guidelines that Ralf Engelschall provided in various
articles and in the documentation of mod_rewrite.  There is an online
article here:
http://www.webtechniques.com/archives/1998/05/engelschall/

The nodes are dual Intel P-III, 512MB RAM, Linux kernel 2.4 in SMP mode
(RedHat 7.2).

I tried building a version of this using Apache 2.0.x (2.0.28 and 2.0.36).
It does work,
however it is incredibly slow.  Benchmarks show that the connections are
happening fast enough, but it seems that something else internally is
stalling, as the response times are very high.  For example, a max
response time for a particular benchmark I ran was 0.2 sec for a static
page in v1.3, the new version has levels of 10 or 20 seconds.
In case it was a problem with mpm-worker, I did another build using mpm-
prefork.  Same poor results with both.  I have tried many different
settings with StartServers, MaxClients, ThreadsPerChild, etc.  No effect.

Is there some known "gotcha" with trying to use 2.0 in this way?  Or has
anyone had good results?  I'd be interested in seeing your httpd.conf and
config.nice files from your build.  I really was hoping to see an
improvement with the mpm-worker module.

Regards,
Robert Jones



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org


Mime
View raw message