The University of British Columbia

Sequencing and assignment-are two important issues we need to address when scheduling real-time tasks on a multiprocessor architecture. Different problem representations can schedule such tasks, each emphasizing either a sequencing or assignment task. A sequence-oriented representation satisfies problem constraints by emphasizing the search for an appropriate tasks order, while an assignment-oriented representation emphasizes the search for an appropriate assignment of tasks. The authors introduce the RT-SADS algorithm, which uses an assignment-oriented representation to dynamically schedule tasks on the processors of a NUMA architecture. The proposed technique automatically controls and allocates the scheduling time to minimize the scheduling overhead and deadline violation of real-time tasks. The authors compare RT-SADS' performance with another dynamic algorithm that uses a sequence-oriented representation. The results show interesting performance tradeoffs among the candidate algorithms.