A formalism for describing and simulating systems with interacting components.
Corr, Glenn A.
MetadataShow full item record
This thesis addresses the problem of descriptive complexity presented by systems involving a high number of interacting components. It investigates the evaluation measure of performability and its application to such systems. A new description and simulation language, ICE and it's application to performability modelling is presented. ICE (Interacting ComponEnts) is based upon an earlier description language which was first proposed for defining reliability problems. ICE is declarative in style and has a limited number of keywords. The ethos in the development of the language has been to provide an intuitive formalism with a powerful descriptive space. The full syntax of the language is presented with discussion as to its philosophy. The implementation of a discrete event simulator using an ICE interface is described, with use being made of examples to illustrate the functionality of the code and the semantics of the language. Random numbers are used to provide the required stochastic behaviour within the simulator. The behaviour of an industry standard generator within the simulator and different methods of number allocation are shown. A new generator is proposed that is a development of a fast hardware shift register generator and is demonstrated to possess good statistical properties and operational speed. For the purpose of providing a rigorous description of the language and clarification of its semantics, a computational model is developed using the formalism of extended coloured Petri nets. This model also gives an indication of the language's descriptive power relative to that of a recognised and well developed technique. Some recognised temporal and structural problems of system event modelling are identified. and ICE solutions given. The growing research area of ATM communication networks is introduced and a sophisticated top down model of an ATM switch presented. This model is simulated and interesting results are given. A generic ICE framework for performability modelling is developed and demonstrated. This is considered as a positive contribution to the general field of performability research.