A Hard Real-Time Static Task Allocation Methodology for Highly-Constrained Message-Passing Environments

This paper presents a method of exploiting the functional parallelism present within a class of well-defined deterministic software systems to achieve real-time execution on a highlyconstrained MIMD message-passing architectural model. The methodology is targeted to software systems that do not respond well to conventional data parallel techniques because of the irregular flow of data resulting from their structure. Such software systems are assumed to be composed of computational and i/o processes (tasks) that execute in a periodic manner under hard real-time constraints. The method performs automatic assignment, mapping, and scheduling of these tasks to the available set of processors assuming a reduced complexity hardware configuration that employs synchronous non-buffered (lock-step) communication between the sending and receiving processors, and simple store-and-forward routing of messages between processors. This methodology is unique in that it combines simulated annealing with a newly developed listbased heuristic to obtain very good solutions in the majority of cases. The effectiveness of the method has been illustrated by applying it to a large number of randomly generated task systems and to a real-world simulation of a U.S. Space Shuttle Main Rocket Engine.