Partitioning Sporadic Task Systems upon Memory-Constrained Multiprocessors

Most prior theoretical research on real-time partitioning algorithms for multiprocessor platforms has focused on ensuring that the cumulative computing requirements of the tasks assigned to each processor does not exceed the processor's processing power. However, computing capacity is often not the only limiting resource: on many multiprocessor platforms each individual computing unit may have limited amounts of multiple additional types of resources (such as local memory) in addition to having limited processing power. We present algorithms for partitioning a collection of sporadic tasks, each characterized by a WCET, a relative deadline, and a period, upon a multiprocessor platform in a manner that is cognizant of such additional constraints as well as the processing capacity constraints.