Customer Classes and Closed Network Models: a Solution Technique

Networks of queues models provide a versatile and powerful means of analyzing and predicting the performance of computer systems. Efficient computational techniques have been developed for solving only a subset of the theoretically solvable models. This paper extends these techniques to cover multi—class models in which customers may change class membership. These extensions provide performance data in terms of the customer classes which were defined in the model. The paper includes a description of the extended solution techniques, bounds on the number of operations and the amount of storage required, and a demonstration of the techniques in solving a model with multiple resources, multiple customer classes, and switching of customers between classes.