Purpose. Multiple processors can be integrated together into a multi-core platform and deal with large-scale science problems because of its excess computation and extensive parallelization. This paper mainly aims at tasks scheduling problem on this platform with shared memory and communication channel. Methodology. Generally, intensive-computation job is portioned into coarse-grain sub-tasks so that tasks are executed in parallel (simultaneously) to improve the computation performance. The proposed algorithm employs constraint programming to find a feasible solution and uses local search to impove the solution and speed up the process. Findings. Sharing a variety of available resources, the multi-processors scheduling is a complex combinatory optimization problem and resource-constrainted problem. We investigate how to obtain a rational solution or sub-optimal solution in a short period. At the same time, tasks with different features are also investigated to find how the performances of applications are influenced in specific target platform. Originality. When tasks are assigned to different processors, various resource constraints, including data storage, executing cost, tasks priority and communication cost among processors, must be considered. A hybrid algorithm based on Tabu search is studied to optimize the completion time of applicaitons satisfying these contraints. Practical value. The algorithm is implemented in IBM ILOG CPLEX Optimization Studio environment and gives better results compared with other algorithms. The proposed algorithm is an effective strategy and the technique can improve search efficiency and solution performance for multiprocessor scheduling problem.

  Якщо, ви не знайшли інформацію про автора(ів) публікації, маєте бажання виправити або відобразити більш докладну інформацію про науковців України запрошуємо заповнити "Анкету науковця"