The equilibrium joint state probabilities of interacting Markov processes are obtained in a hierarchical way, by finding the reversed process of the interaction in terms of the reversed processes of the components. From a reversed process, a product-form solution for the joint state probabilities follows directly. The method uses a Markovian process algebra formalism and generalizes the recent Reversed Compound Agent Theorem (RCAT) to solve a diverse class of concurrent systems. This class includes processes with shared, exclusive resources, a customer-oriented specification of a last-come-first-served (LCFS) queue with Coxian service times and an extended PS queue with a non-product form solution. From these results, a new, very short proof of the BCMP theorem ensues. The principal advantage of the methodology is its potential for mechanisation and symbolic implementation. Indeed, many non-standard product-forms have emerged directly from the compositional approach.
Information from pubs.doc.ic.ac.uk/rcat2-nsmc.