MODELLING DYNAMIC DISTRIBUTED SYSTEMS.

The authors present a model of structure and communication within distributed systems of concurrently executing components (agents) communicating by asynchronous message-passing, and formulate a definition of observational equivalence between such systems. The model accomodates the dynamic system connectivity required to model open and continuously operating systems. New communication links can be created; an end-point of a communication link (a port) possessed by one agent can be sent, as part of a message, to another agent, which thus dynamically acquires the ability to communicate on that link. Ports can be imported from and exported to the system's external environment, thus allowing changing connectivity of the system as the whole. The authors introduce a notion of observational equivalence such that two systems are defined as being observationally equivalent if and only if replacing one by the other in any larger system does not affect the operational behavior of the remainder of that larger system (that remainder being an observer which thus cannot detect the difference).