The load and performance test tool (mainly for Web-sites) Apache JMeter
(http://jakarta.apache.org/jmeter/) doesn't support the definition of
probabilistic user behavior, i.e. given a currently visited Web site, the
next site to be requested is chosen based on configurable probability values.
JMeter solely provides a random controller with uniformly distributed
probabilities.

A Markov Chain is a formal model having a number of states and transitions.
Transitions between states are annotated with probabilities. The sum of
probabilities of outgoing transitions must be 1. 

We added a new Controller to JMeter, supporting the test plan execution
based on probabilistic user behavior(s) itself based on the Markov chain
concept (as well as concepts known from finite state machines, e.g. guards
and actions). Moreover, our extension enables the variation of the number of
active parallel user sessions (threads) during test run, to be specified by
user-definable mathematic formulae. 

###
# Install Markov4JMeter
###

see tutorial "doc/tutorial.pdf" from the binary distribution

###
# Build Markov4JMeter
###
To build Markov4JMeter you need to perform the following steps:

1.) Download the source archive
2.) Call 'ant'

Remark:
The lib/opt directory contains the following Jars needed for compilation only:
- ApacheJMeter_core.jar 
- jorphan.jar
- logkit-<version>.jar
- js-<version>.jar
These Jars were taken from the most recent JMeter release archive.