Interactive Modeling of Computer Systems

paper provides a nontechnicul introduction to queuing network concepts, a short introduction to results of research into analytical modeling methodology, and u set of simple but typical examples that illustrute the upplication of that methodology to problems of computer perfbrmunce. The objectives of this paper are to introduce concepts of the theory of queuing networks, to describe an APL program-called QNET4-for the solution of queuing network problems, and to give examples of the analysis of computer systems using the methods of QNET~. In the process of constructing models, we wish to clearly distinguish between modeling and methods of solution. Modeling is the process of mapping a real system into a suitable abstract representation, such as an equation or a queuing diagram. It is the modeler who decides those aspects of a system that are sufficiently important to be represented. Modeling, especially on a high level of abstraction, is a difficult endeavor. Much remains to be done to establish a proven modeling methodology. Methods of solutions are necessary to provide numerical results. A method of solution is usually capable of solving a certain class of models. (Such a class of models is sometimes called an abstract model.) The abstract model of this paper is the network of queues. Algebraic and simulation methods are two important methods of solution of such models. The APL program discussed in this paper, which we call Q N E T ~ , is an implementation of the algebraic method of solution. Note that we view simulation as a method of solution, not as a model class that is distinct from the class of algebraic (or analytic) models. Queues and queuing networks The term queue denotes an individual resource of the queuing network. Several queues are connected into a queuing network by " routing, " which is described later in this paper. A queue consists of the following three elements: a waiting room, a set of servers or processors, and a queue discipline, i.e., a rule for determining the scheduling of service to jobs.