Multiprocessor Task Scheduling with Single Resource Constraints

Abstract

An important problem arising in a multiprocessor system context is that of task allocation (scheduling) among a set of processors working in parallel. One of the assumptions commonly imposed on the processor scheduling theory is that each task is processed on at most one processor at a time. In fact, all polynomial-in-time algorithms as well as NP-completeness results for task scheduling were obtained by this assumption.

However, in recent years, together with the rapid development of microprocessor and especially multi-microprocessor systems, the above assumption has ceased to be justified in some important applications. This is especially true with some highly specialized systems where tasks may require more than one processor simultaneously. These problems create a new direction in the processor scheduling theory, in which preliminary results concerning the pre-emptive scheduling of tasks requiring one or k (k fixed, > 1) processors were obtained in [1] and [2] for the schedule length criterion.