Simulations between specifications of distributed systems

Abstract

In the stepwise development of a distributed system, the problem arises of verifying that a specification at a lower level of abstraction correctly implements a specification at a higher level of abstraction. Forward and backward simulation have been proposed as verification techniques for this problem. In this paper, we study forward and backward simulation in a framework where specifications are given as labeled transition systems with fairness requirements. We aim at clarifying the connection between simulations and the auxiliary variable constructions of Abadi and Lamport. In the paper, we also relax the earlier restriction that backward simulations be finitary. For a simple specification notation, similar to the action system formalism or Unity, we furthermore present proof rules that correspond to forward and backward simulations. Finally, we relate the forward and backward simulation techniques to subset-constructions that can be used in automata theory, e.g. for deciding language containment.