Distributed synthesis of asynchronously communicating distributed process models

Abstract

We investigate to what extent existing algorithms for the discovery of component models from event logs can be leveraged to a system of asynchronously communicating components. Here, Enterprise nets model local processes, while Industry nets are compositions of Enterprise nets which interact through asynchronous message passing. We investigate the relation between the behaviour of an Industry net and that of its constituting Enterprise nets and we formalise the (causal) structure of global (Industry net) behaviour in terms of a partial order derived from the message passing. Next, we specify how (existing) algorithms for the discovery of isolated processes, can be adapted to enable the discovery of Enterprise nets, and we demonstrate how to combine these Enterprise nets into an Industry net. Using the results on the structure of the global behaviour, we relate the behaviour of the Industry net thus synthesised to the behaviour of the Enterprise nets and show how fitness of the Enterprise nets (the event log provided as input is included in the behaviour of the discovered net) is preserved as fitness of the Industry net. Moreover, we discuss possible underfitting of the global model (the model exhibits more behaviour than observed in the event log) and show how it can be explained in terms of concurrency between the component models and a completeness property of the event log.