httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@worldgate.com>
Subject Re: web benchmarking
Date Wed, 18 Jun 1997 16:21:35 GMT
On Wed, 18 Jun 1997, Dean Gaudet wrote:

> I thought zeusbench looked good from a very brief perusing. 
> www.zeus.co.uk I think.  Emphasis on brief. 

Looked at it and use it for basic testing, but it is quite limited and I'm
not sure it is really a good benchmark.  It is not threaded and only uses
one process; it seems to me like this design gives Zeus's web server
(which is also not threaded and only uses one process) a potential
advantage by not exposing any of the design tradeoffs involved in
doing things that way.

Oh, it is also horrible unforgiving of webservers that start refusing or
dropping connections when they can't handle the load; with some crappy
servers, I have to restrict the number of clients to <50 or zb will just
die completely with a broken pipe (I'm assuming SIGPIPE) error...

> 
> Dean
> 
> On Tue, 17 Jun 1997, Marc Slemko wrote:
> 
> > anyone have any comments on programs for web benchmarking?  I keep saying
> > that I am planning to do some real benchmarking, comparing both Apache
> > against other Unix servers and against NT servers.
> > 
> > I have a server I can do whatever with (PPRO/180 w/96 megs), a couple of
> > Pentium Unix boxes I can pull to act as clients, enough 100 meg cards
> > (although no spare 100 meg hub...).   I don't have a network of Windows
> > machines to act as clients like some benchmarks want.
> > 
> > I think I will end up using webstone, but all the scripts it uses are ugly
> > things designed for SGI boxes where everything is insecure.  
> > 
> > I looked at specweb, but aside from $800 for purchase it just doesn't look
> > to be that great.
> > 
> > 
> > 
> 


Mime
View raw message