[mnw.pm] Stress testing?

[elaine ashton]

> Lisa Nyman (GEO-CSB) wrote:
> > Does anyone know of a script/module/etc archived somewhere that deals with
> > stress testing a server and application - sort of bombarding the server
> > with requests to see how it handles the presure?
> 
> if you're talking about a -web- server and a CGI
> script, you may want to look at Lincoln Stein's 
> torture.pl, from TPJ #8, or:
> 
> http://orwant.www.media.mit.edu/tpj/programs/Issue_08_CGI/torture.pl

Actually, getting an accurate 'stress' test of a web server is a lot more
difficult than it would appear at first. I had to beat up an ultra 5 and
an ultra 10 for BBN at one point as they wanted a low end web server and
wondered if it could fit the bill. I had an old C program that would go
out and get one doc and return the stats after 25k hits. This was fine but
it was only for one document. Then I made another out of Perl that took a
list of urls and randomised them on the get so that it would seem more
'natural'. Then we moved on to SpecWEB which supposedly emulaates real
user stress on the machine. Lincoln's torture.pl works well for testing
cgi. I have used it before and usually recommend it to web site
developers. 

Lisa, if you are stress testing Solaris and Netscape enterprise server let
me know as there are a lot of things you can do on the server/ns side to
tune up performance as well. 

e.