Development of a quasi-birth-and-death sensitivity analysis tool

Abstract

In the past, the performance analysis of random systems has been of great interest. Such an analysis often consists in the computation of so-called performance measures. Sometimes, it is desired to evaluate the sensitivity of the performance with respect to variation applied to the parameters of the considered system. Such an evaluation is usually referred to as sensitivity analysis.
Markov processes are intensively used to model the evolution of random systems over time. We are especially interested in a particular case of Markov processes, called the Quasi-Birth-and-Death process (QBD). The
QBDs combine a nice modelling expressiveness with the possibility of using efficient evaluation methods.
We aim the development of a tool that can perform sensitivity analyses on QBDs. More precisely, the tool is required to provide methods to specify a QBD process, procedures to carry out the analyses, and interfaces to display the produced results. To achieve this, we present the theoretical foundations of methods to compute the performance measures, as well as the tool and a case study made with it.

Date of Award	2011
Original language	English
Awarding Institution	University of Namur
Supervisor	Marie-Ange Remiche (Supervisor)