Bounding the end-to-end response time in multiprocessor real-time systems

In a multiprocessor real-time system, a task may execute in turn on several processors before it completes. The task may have different priorities on different processors and may execute on some processors more than once. Existing techniques for bounding the response times of such tasks are not effective. In this paper, we present a method to compute tighter upper bounds on their response times.<
>