Implementation relations and probabilistic schedulers in the distributed test architecture

Abstract

We present a complete framework to formally test systems with distributed ports where some choices are probabilistically quantified while other choices are non-deterministic. We define different implementation relations, that is, relations that state what it means for a system to be a valid implementation of a specification. We also study how these relate. In order to define these implementation relations we use probabilistic schedulers, a more powerful version, including probabilistic choices, of a notion of scheduler introduced in our previous work. Probabilistic schedulers, when applied to either a specification or an implementation, resolve all the possible non-determinism, so that we can compare purely probabilistic systems.