Energy-Efficient Scheduling for Real-Time Tasks on Uniform Multiprocessors

Our aim in the research is to address the scheduling problem of a uniform multiprocessor platform with a periodic task set. The processors in the platform have the same instruction architecture. Besides, each processor has a limited range of discrete available speeds and the speed of each processor can be adjusted independently from each other. We propose an offline task-to-processor assignment algorithm, the Largest Task First with Available Speed Considerations (LTFwAS) algorithm to consider discrete available speeds for the processors. The algorithm can derive a feasible task assignment with the minimal energy consumption and has the time complexity of O(MNX), where M, N, and X are the numbers of uniform processors, tasks, and available speeds, respectively. A series of experiments are conducted to evaluate the proposed algorithm. The experimental results demonstrate that the performance of the proposed LTFwAS algorithm is greatly similar to that of the optimal algorithm.