考虑处理机下线时间的可分任务调度优化模型

发布时间：2018-04-29 04:03

本文选题：可分任务调度 + 遗传算法　；参考：《电子科技大学学报》2017年01期

【摘要】：随着科学应用逐渐趋于数据密集型计算,为并行与分布式系统寻求高效的任务调度策略成了研究的热点问题。已有的可分任务调度模型均假设所有处理机都能100%的完成子任务的计算,即处理机在完成任务计算之前一直保持在线状态。实际上,并行与分布式系统中不同处理机的在线时间可能不同。若忽略处理机的在线时间,为其分配的任务量过大,则任务的完成时间可能超出处理机的下线时间,从而造成任务的计算无法按时完成。因此,为处理机分配任务时应充分考虑处理机下线时间的限制。为解决上述问题,该文提出了一种新的考虑处理机下线时间的可分任务调度优化模型,并设计了全局优化遗传算法求解该模型。最后,通过仿真实验结果验证了模型和算法的有效性。
[Abstract]:With the increasing trend of scientific applications to data-intensive computing, finding efficient task scheduling strategies for parallel and distributed systems has become a hot issue. The existing divisible task scheduling models assume that all processors can complete 100% of the calculation of subtasks, that is, the processor remains online until the task calculation is completed. In fact, different processors in parallel and distributed systems may have different online times. If the online time of the processor is ignored and the amount of tasks assigned to the processor is too large, the completion time of the task may exceed the off-line time of the processor, thus the calculation of the task cannot be completed on time. Therefore, when assigning tasks to the processor, the limitation of the downline time of the processor should be fully taken into account. In order to solve the above problems, this paper presents a new scheduling optimization model for divisible tasks considering the downtime of the processor, and designs a global optimization genetic algorithm to solve the model. Finally, the effectiveness of the model and the algorithm is verified by the simulation results.
【作者单位】：西安电子科技大学计算机学院;
【基金】：国家自然科学基金(61402350,61472297,61572391) 中央高校基本科研业务费专项资金(JB150307)
【分类号】：TP18

【相似文献】