Due to its potential for high reliability and high performance, a carefully designed distributed system can make an ideal controller for real-time applications. However, the main drawback of a distributed system is that the communication latency between tasks executing on two different nodes can be fairly large. This research is to improve the communication performance by addressing the problems related to message passing in a distributed system with point-to-point interconnection topology. First, the assignment of deadlines to messages based on the deadlines of related tasks is formalized. Message passing techniques are then developed that have higher probabilities of messages meeting their assigned deadlines. Since, by the first objective, the deadlines have been assigned in a proper manner, the new message passing techniques result in a better communication mechanism between the real-time tasks. Our approach is to first characterize the probability distribution function of message delivery times. Using this distribution, the deadlines for messages from the deadlines of related tasks will then be derived. Finally, use will be made of the multiple disjoint paths in a distributed system to increase the probability of a message meeting its assigned deadline.
