Generalized Stochastic Petri Nets in Performance Evaluation for Queueing Networks Mihaela-Hanako Matcovschi and Cristian Mahulea* *Department of Automatic Control and Industrial Informatics Technical University “Gh. Asachi” of Iasi, Blvd. Mangeron 53A, 6600 Iasi, Romania Phone: +40-32-230751 , Fax: +40-32-214290 E-Mail: {mhanako, cmahulea}@delta.ac.tuiasi.ro Abstract: Markovian queueing networks may be modeled by generalized stochastic Petri nets. The simulation of a model in a properly chosen simulator provides information on the performances of the actual system, mainly in case that the analytical study is intricate. After presenting some general notions on queueing networks and on generalized stochastic Petri nets, this paper points out the PN Toolbox for MATLAB, an instrument for modeling and simulating several types of Petri nets. The comparison of results obtained through simulation to the analytical ones for a certain model prove the usefulness of this software package for performance evaluation. Keywords: queueing network, performance evaluation, stochastic Petri nets, MATLAB. 1. Introduction There are three general techniques for predicting a system’s performances: direct measurement, analytical modeling and simulation. Extensive measurement is most accurate but it is expensive and cannot be used in all situations encountered during the lifetime of a system. Due to the increase of the complexity and automation encountered in modern industrial systems, the performance evaluation for such systems relies more and more on simulation rather than on analytical methods. Moreover, the growing interest in simulation-based techniques for the prediction of reliability, availability and combined performances of complex systems has been stimulated by the advent of fault-tolerance concept. Such techniques provide manufacturers cheap and versatile tools, useful in all the phases of system life cycle. The goal of ease of use is best met with software packages that provide an easy to use and powerful user interface for model specification. Petri nets (PNs) proved to be a powerful graphical and mathematical modeling tool that allows describing and analyzing complex systems exhibiting concurrency, synchronization and conflicts such as computer systems, communication systems, and manufacturing systems. Their ability to model timed and probabilistic behavior is essential in the field of performance evaluation. This need lead to various different extensions of the PN formalism, where the class of Stochastic Petri nets (SPNs) gained the widest acceptance. Many a package software for supporting engineers in the definition and analysis of such type of models have been built over the last years, such as SPNP, GreatSPN, TimeNET. The choice of a particular tool is mainly based on its accessibility as well as on the type of models under study. Our paper focuses on performance evaluation for queueing networks based on SPN models and simulation procedures. Due to our experience in exploiting MATLAB numerical and programming capabilities, we have designed and implemented a new software, namely PN Toolbox for MATLAB, devoted to modeling, simulation and analysis of PN type models. This paper has the following structure. Section 2 describes the basics of queueing networks and of their performance evaluation. Generalized stochastic Petri nets (GSPN) used for modeling such complex systems are briefly put forward in Section 3. A concise representation of the facilities available in PN Toolbox is realized in Section 4. An example of its usage in analyzing queueing networks are further on delivered in Section 5. The concluding remarks in Section 6 point out the main results of this paper.