httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ben Laurie <...@gonzo.ben.algroup.co.uk>
Subject Re: Real World Webserver Performance
Date Wed, 14 Feb 1996 17:58:16 GMT
> 
> >
> > it can't work. There's no mention of the type of requests handled by
> > the server (plain HTML, CPU or I/O intensive CGI, large mpegs...) and 
> > no mention of the type of user (mostly local, mostly modem users, mostly
> > of a shitty sprintlink line....)
> > 
> > There are too many variables IMO.
> >
> 
> 
> There's only one way to compare performance and that's to run the server
> on the exact same physical machine and serving the exact same sequence of
> requests with the exact same processes running at the same time.
> 
> Easily doable really. You just have a benchmark box on which only the
> server and necessary processes to keep the server working are running and
> then you develop a benchmark program that generates those requests which
> you want performance measurements for. It's basically no different to
> how you'd write and use something like iozone.
> 

The trouble with benchmarks is that it isn't possible to construct a benchmark
which models my exact mix of hits. Or even one which will predict performance
under my exact mix.

Also, how do you account for different h/w and OS? You end up with an
impossible number of tests.

One result that could be easily obtained from a survey would be upper limits
on performance for each platform, a useful item in itself. I'm sure there are
others.

Cheers,

Ben.

-- 
Ben Laurie                  Phone: +44 (181) 994 6435
Freelance Consultant        Fax:   +44 (181) 994 6472
and Technical Director      Email: ben@algroup.co.uk
A.L. Digital Ltd,           URL: http://www.algroup.co.uk
London, England.

Mime
View raw message