The tool we use is not cheap. I think we paid like $35K or something for it.
Its called E-Test Suite, from RSW. We did buy several licenses, so its
possible just buyingthe e-load program by itself is feasible..however, you
do need the e-test program to create scripts, so I imagine its still not all
that cheap.
Unlike some programs, e-test is very easy to use. You almost don't need the
manual. You can use simple text files for databanks (to allow multiple users
at the same time with different information), it validates the pages, and
its as simple as pressing a record button, typing in the url, and
clicking/typing away on the url you went to. Then you stop, save the script,
and run it. Later, you can use a databank then run it in the load test which
can simulate up to 100 users at the same time, or more if you buy licenses
in 100-virtual users groups. The thing is..if your site can handle 100
simultaneous users all submitting, sending, etc..then most likely it can
handle 1000 at the same time..because most of the time a person is not
actually submitting..they are filling out forms, reading, etc. Also, keep in
mind that transactions are heavier on the system than simple page requests
(navigation via menus, links, etc that don't use the database..just go to
another page). So it really depends on what you are testing.
I posted a test of Orion application server running under load of 25 users
on a simple login, and on my PIII650 at work with 512MB RAM and ATA 66 HD,
using our database over the network, I was able to achieve about 4.5million
pages per day requests. That is off of one single "slow" (by todays
standards) cheap built-up pc and an Oracle database on a E450 Server with
512MB RAM, 1 cpu and an 18GB SCSI drive with 15 connections available in the
pool. 25 users isn't much, but for our site, 25 simultaneous users would be
a lot right now. It all depends on the site as well..each of our users may
send orders in the millions through our site, so while our actual volume of
people on the site is low, we generate high-value transactions. On the flip
side, you can probably load-test just the speed of Orion returning JSP
pages. I had done so, and will post some results when I get a chance.
> -----Original Message-----
> From: Robert Krueger [mailto:[EMAIL PROTECTED]]
> Sent: Saturday, October 28, 2000 2:25 AM
> To: Orion-Interest
> Subject: load testing tools/procedures
>
>
>
> hi,
>
> what are people out there using for load testing when they've
> finished a
> web-application? the (affordable) solutions I can think of is JMeter
> (java.apache.org) or simulating user sessions by coding tests with a
> package like HttpUnit. I'm particularly interested in testing
> applications
> which use sessions and need the simulation of real users
> (i.e. log in -
> view catalog - modify shopping basket - submit order - logout
> kind of stuff).
>
> any thoughts on this, anyone?
>
> regards,
>
> robert
> (-) Robert Kr�ger
> (-) SIGNAL 7 Gesellschaft f�r Informationstechnologie mbH
> (-) Br�der-Knau�-Str. 79 - 64285 Darmstadt,
> (-) Tel: 06151 665401, Fax: 06151 665373
> (-) [EMAIL PROTECTED], www.signal7.de
>
