Performance analysis of timed Petri nets by decomposition of the state space

Thesis (M. Sc.), Memorial University of Newfoundland, 1998. Computer Science Bibliography: leaves 63-67. Performance evaluation of systems is a very important part of system design. Modeling tools which allow the analysis of systems and their behaviors should also provide performance analysis of the...

Full description

Bibliographic Details
Main Author: Suciu, Floarea, 1969-
Other Authors: Memorial University of Newfoundland. Dept. of Computer Science
Format: Thesis
Language:English
Published: 1998
Subjects:
Online Access:http://collections.mun.ca/cdm/ref/collection/theses4/id/1862
Description
Summary:Thesis (M. Sc.), Memorial University of Newfoundland, 1998. Computer Science Bibliography: leaves 63-67. Performance evaluation of systems is a very important part of system design. Modeling tools which allow the analysis of systems and their behaviors should also provide performance analysis of the modeled system as it is less costly to perform changes at the model level. Petri nets are becoming popular modeling tools that can represent and analyze concurrency, parallelism, synchronization, mutual exclusion and conflicts. However, time and space requirements of the classical approach of exhaustive generation of all possible behaviors of the system (or its state space) grow quickly with the size of the model. An alternative approach based on structural properties can be applied only to particular classes of nets. -- A new way to derive performance measures for timed Petri nets is based on decomposition of the state space. This is a hybrid method that uses both reachability and structural analysis. The state space of the original net is decomposed into state spaces of smaller nets, and these smaller nets are then analyzed by the reachability analysis method. Since the nets are quite simple, reachability analysis is straightforward and cannot be affected by the "state explosion" problem. The performance indices for smaller nets are then used for performance analysis of the original net.