GSPN Models: Sensitivity analysis and applications
Sensitivity analysis of continuous time Markov chains has been considered recently by several researchers. This is very useful in performing bottleneck analysis and optimization on systems especially during the design stage. However the construction of these large and complex Markov models is tedious and error-prone process. Generalized Stochastic Petri Nets (GSPN) provide a very useful high-level interface for the automatic generation of the underlying Markov chain. This paper extends parametric sensitivity analysis to GSPN models. The rates and probabilities of the transitions of GSPN models are defined as functions of an independent variable. Equations for the sensitivity analysis of steady-state and transient measures of GSPN and GSPN reward models are developed and implemented in a software package. An example illustrating the use of sensitivity analysis is presented.
