From pac@fortuitous.com  Fri May 12 04:41:33 2006
From: pac@fortuitous.com (Phil Carinhas)
Date: Thu, 11 May 2006 22:41:33 -0500
Subject: [Infrastructures] Testing general system functionality via user simulation
In-Reply-To: <442D87EC.8010401@tufts.edu>
References: <442D87EC.8010401@tufts.edu>
Message-ID: <20060512034133.GA28341@mail.fortuitous.com>

On Fri, Mar 31, 2006 at 02:50:04PM -0500, Marc Chiarini (Tufts) wrote:
> Hello All,
> 
> I am trying to analyze program dependencies in Linux.  Does anyone know 
> of any suite/testbed/infrastructure that would allow me to create 
> "automated virtual users" that would run on real (or virtual systems), 
> exercising various large subsets of software in a reasonable, if not 
> realistic fashion?  I'm not sure that simulation of the systems and 
> users will get me what I want.  I think I need something that sits on  a 
> machine and pretends to be  something like a user, running various 
> applications in different classes (I will settle for non-GUI ;).  There 
> needs to be at least some minor variability to these virtual users' 
> behaviors.  It is not important to my research that they be anywhere 
> near realistic, as long as they are executing commands with reasonable 
> parameters.
 
   Have you tried or thought of XEN virtual servers? It may not be exaclty
what you are looking for, but it could work.
http://www.cl.cam.ac.uk/Research/SRG/netos/xen/

> Another way that I might accomplish part of my goal is to utilize a test 
> suite that runs (with a variety, but obviously not all parameters) as 
> many non-root commands on a system as possible.  It seems that autoconf 
> provides a limited infrastructure for doing something like this, but I 
> haven't investigated it thoroughly.
 
  Another approach may be to do analytical modeling of the systems in
question. This will only give you statistical and numerical information 
about the system you are designing. In many cases, its much easier to
model it and get out the stats, than to do a full blown simulation.
Modeling has the extra benefit of being easily expandable.

 A combination aproach might also get you the answers you are looking for 
 as well.

-Phil Carinhas
 http://fortuitous.com/en/services/
--
 .--------------------------------------------------------.
 | Dr. Philip A. Carinhas      || http://fortuitous.com   | 
 | Fortuitous Technologies Inc || Tel : 1-512-351-7783    | 
 | Performance Engineering, Capacity Planning & Unix Svcs |
 `--------------------------------------------------------'