Discrete Event Simulation in the Möbius Modeling Framework

Discrete event simulation is an important method to determine the performance and dependability of many systems, including computer and communication networks. Whereas many analytic solution methods require models to conform to certain criteria or produce state spaces no larger than a computer system can manage, simulation has none of these limitations. This is the primary reason that simulation remains such a popular solution method despite the development of new analytic methods. There are many modeling packages available, most of which include simulators. However, each of these packages is able to create and solve models only within a single formalism. This is extremely limiting both to model construction and solution, but also to the advancement of new formalisms and solution engines. The Möbius project solves these problems by providing an object-oriented, formalism-independent modeling environment. By utilizing the abstraction of this design, we are able to create a fast, efficient, cross-formalism simulation engine that provides solution power unavailable in other packages. This simulator is able to make use of distributed computing networks to manage multiple processors simultaneously. Also available is an assortment of built-in execution policies designed to simplify common modeling tasks and provide added modeling functionality. iv To my family and friends. v ACKNOWLEDGMENTS There are many people I would like to thank for their support both while I was here at the University of Illinois and throughout my academic career. First, I would like to thank Dr. William H. Sanders both for his technical guidance and for the opportunity to work with the Performability Engineering Research Group. I would also specifically like to thank Jay for their friendship and support throughout the design and development of the Möbius project. Thanks also to Dan Deavours and Doug Obal for allowing me to tap their vast knowledge of system modeling and for their continual help and friendship. Thanks also to Jenny Applequist for her help in reviewing this thesis. Systems Technology Group, including Renee Langefels and Peter Alejandro, for their long term funding of the UltraSAN and Möbius projects. I would especially like to thank my family for their constant guidance and loving support. Thanks also to all my friends both here and abroad and to all those who have helped me along the way.