You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by Subramaniam <nb...@gmail.com> on 2009/03/15 07:53:53 UTC

RE: Markov4JMeter: JMeter extension for probabilistic user behavior (Markov chains)

Hi Mark,

   I'm Subramaniam from India. I'm trying to learn more about representing
user behavior model in the form of CSV files.

   Suppose  we have a web application having a home page, The home page has
a link to a sellers page and another link to buyers page.

        State 1 : Home page
        State 2 : Sellers page
        State 3 : Buyers page

 Under the context of this application, If we assume that a buyer would
never sell and a seller would never buy.  Every buyer would certainly buy an
item and every seller would certainly sell an item.We can have the following
two behaviors

  Sellers behavior
  --------------------
                       Home Page   Sellers Page    Buyers Page    $
  Home Page           0                1                  0              0
  Sellers Page         0                 0                  0              1
  Buyers Page         0                 0                  0              0



  Buyers behavior
  --------------------
                       Home Page   Sellers Page    Buyers Page    $
  Home Page           0                0                  1              0
  Sellers Page         0                 0                  0              0
  Buyers Page         0                 0                  0              1

 In a sample of 10 users , if we always have 6 buyers and 4 sellers, The
relative frequency for representing the behavior mix would  be

Behavior mix
--------------
 Seller   0.4
 Buyer   0.6


My doubt
-------------

Is it necessary that the behavior matrix must certainly be an n X (n+1)
matrix.
In the above example, Seller row for a buyer and buyer row for a seller are
not necessary, can we eliminate these rows in the csv file.

  Thanks,
Subramaniam.










Mark McWhinney wrote:
> 
> It works great.
> 
> Please keep us posted on further updates.
> 
> 
> -----Original Message-----
> From: Mark McWhinney [mailto:msm@portata.com] 
> Sent: Tuesday, June 05, 2007 4:48 PM
> To: 'JMeter Users List'
> Subject: RE: Markov4JMeter: JMeter extension for probabilistic user
> behavior
> (Markov chains)
> 
> Hello André,
> 
> That is very timely.  I was just about to post a message about how to set
> up
> a Markov style load model in JMeter using the various types of controllers
> and branching logic.
> 
> I will spend tomorrow setting up and using your controllers.
> 
> Mark
> 
> 
> -----Original Message-----
> From: Andre van Hoorn [mailto:andre.van.hoorn@Informatik.Uni-Oldenburg.DE] 
> Sent: Tuesday, June 05, 2007 7:18 AM
> To: jmeter-user@jakarta.apache.org
> Subject: Markov4JMeter: JMeter extension for probabilistic user behavior
> (Markov chains)
> 
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hi,
> 
> 
> in order to generate workload based on probabilistic user behavior
> models (Markov chains) with a varying number of simulated users, I
> developed an extension for JMeter named Markov4JMeter.
> 
> It adds two new logic controllers ("Markov Session Controller", "Markov
> State") representing a session model for an application and its states.
> Transitions between states can then be assigned  guards (conditions) and
> actions (e.g. variable assignments) known from state machines. The
> transition probabilities are specified within external behavior model
> files (CSV files containing the transition matrix; templates can be
> generated conveniently). A mix of behavior models with associated
> relative frequencies of their occurence during a test execution can be
> specified.
> 
> Moreover, the number of threads can be varied based on mathematical
> formulae e.g. specified using a BeanShell script.
> 
> I implemented Markov4JMeter as a JMeter component (Jar file) which is
> installed by simply adding it to JMeter's lib/ext folder (it requires at
> least JRE 1.5). It is licensed under the Apache 2.0 license.
> 
> The release (binary and source) including a tutorial (for getting a
> first idea) can be downloaded from http://markov4jmeter.sourceforge.net/
> 
> I would be glad if this extension is helpful to someone else. Feedback
> is greatly appreciated!
> 
> Best regards,
> 
> André
> 
> - --
>  André van Hoorn
>  Birkenweg 21
>  D-26725 Emden
> 
>  Telefon:  +49 (4921) 954231
>  E-Mail:   Andre.van.Hoorn@Informatik.Uni-Oldenburg.DE
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
> 
> iD8DBQFGZXB/LKwmE5jzZmkRAjecAKCwN51wUbEylqybIAQoQ45r0INKkQCcDjkA
> TM0iVERrqp0Yi/7boJyP2UQ=
> =oaBU
> -----END PGP SIGNATURE-----
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> 
> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> 
> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> 
> 
> 

-- 
View this message in context: http://www.nabble.com/Markov4JMeter%3A-JMeter-extension-for-probabilistic-user-behavior-%28Markov-chains%29-tp10970366p22520545.html
Sent from the JMeter - User mailing list archive at Nabble.com.


---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org