Message passing architectures for stochastic and dynamic distributed real time systems

Distributed real time systems have recently evolved that are based on commodity computers and networking technologies. In these systems, critical system resources and functionality are distributed over many different computers and networks. The operational modes of these systems include the integration of heterogeneous event driven information flows from multiple sensor platforms that dynamically change with the operational scenario. The dynamic and stochastic nature of these systems renders allocating the distributed resources based on a-priori worst case loading and capacity scenarios impossible. This paper first discusses the operational characteristics of these systems that effect the requirements for a messaging based communications system. Appropriate message passing protocols are then discussed. Finally, we conclude with a discussion of the architectural support necessary for efficiently implementing a message passing protocol in a dynamic and stochastic distributed real time system.