A major challenge for the immediate future in the field of distributed computing systems is to realize their potential benefits without incurring intolerable penalties in performance, complexity, and cost. There is therefore a substantial interest in establishing sound theoretical foundations for the modeling, design, construction, and reliable maintenance of these systems. Analytic modeling for performance evaluation of distributed systems typically requires the solution of large multichain queueing network models. The size of these large models precludes any use of exact solution techniques. Thus, it is important to develop accurate and cost effective algorithms for the approximate solution of large multichain queueing networks. To meet the challenges of future complex distributed computing systems, approximation techniques will have to be employed.
The work presented in this dissertation was motivated by our interest in analyzing and solving queueing network models of large distributed systems. This dissertation focuses on the development of accurate and cost effective algorithms for the approximate solution of large multichain product form queueing networks, and in particular those that represent the models of large distributed computing systems. The results obtained in this dissertation are applied to a variety of configuration design issues and other distributed systems problems. The work presented here can be used as a platform upon which future queueing networks modeling tools can be constructed.