Protocol engineering from Estelle specifications

Wheeler, Graham (1993)

Bibliography: leaves 129-132.


The design of efficient, reliable communication protocols has long been an area of active research in computer science and engineering, and will remain so while the technology continues to evolve, and information becomes increasingly distributed. This thesis examines the problem of predicting . the performance of a multi-layered protocol system directly from formal specifications in the ISO specification language Estelle, a general-purpose Pascal-based language with support for concurrent processes in the form of communicating extended finite-state machines. The thesis begins with an overview of protocol engineering, and a discusses the areas of performance evaluation and protocol specification. Important parts of the mathematics of discrete-time semi-Markov processes are presented to assist in understanding the approaches to performance evaluation described later. Not much work has been done to date in the area of performance prediction from specifications. The idea was first mooted by Rudin, who illustrated it with a simple model based on the global state reachability graph of a set of synchronous communicating FSMs. About the same time Kritzinger proposed a closed multiclass queueing model. Both of these approaches are described, and their respective strengths and weaknesses pointed out. Two new methods are then presented. They have been implemented as part of an Estelle-based CASE tool, the Protocol Engineering Workbench (PE!V). In the first approach, we show how discrete-time semi-Markov chain models can be derived from meta-executions of Estelle specifications, and consider ways of using these models predictively. The second approach uses a structure similar to a global-state graph. Many of the limitations of Rudin's approach are overcome, and our technique produces highly accurate performance predictions. The PEW is also described in some detail, and its use in performance evaluation illustrated with some examples. The thesis concludes with a discussion of the strengths and weaknesses of the new methods, and possible ways of improving them.