并行异构系统中的一种高效任务调度算法

An Effective Task Scheduling Algorithm for Heterogeneous Parallel Systems

并行分布计算是当前计算机科学的热点之一。调度算法是影响分布式计算的关键因素,也是一个具有挑战性的课题。调度算法是将有通信关系的任务按顺序分配给不同的处理器。提出了一种基于区间插入和任务复制的高效启发式算法,通过对各种随机任务图和高斯迭代任务图进行模拟,与已有算法相比,新算法的效率有很大的提高。

Efficient appl ication scheduling is critical for achieving high performance in heterogeneous computing environment. This paper proposes a novel algorithm based on list-scheduling and task-duplication on a great number of fully connected homogeneous machines. The comparison study, based on both randomly generated graphs and the graphs of Gaussian elimination, shows that the scheduling algorithm significantly surpass existing algorithms.