Network Models for the Conversational Design of Stochastic Service Systems

Numerical analysis of Markovian queueing networks, and graphical communication of problem statements and results, offer a potential for truly conversational use of computers for "high-traf fi"." design of largescale systems. However, one must develop a translator which converts the pictorial language of queueing network diagrams to the data structures required for efficient numerical analysis. This requires the development of new mathematical models for networks, and for the meaning of their components. Such mathematical models are explored, and their role in the development of adequate programming systems is described. These models, algebraic in form, provide a vehicle for the information conveyed directly by the diagram, the information implicit in the symbology of the diagram, the information upon which the actual calculations are performed, and the procedure which transforms the information from the form of the diagram to the form for calculation.