异构集群系统的可分负载多轮调度算法

Multi-round algorithm for scheduling divisible workloads on heterogeneous cluster computing systems

针对更实际的异构集群计算环境,充分考虑处理机具有不同的计算速度、通信能力和存储容量的特性,通过允许计算和通信操作重叠执行,采取多次并行分配计算任务的方法,设计一种可分负载多轮调度算法。实验结果表明,该算法不但能获得与均匀多轮调度(UMR)算法相当的渐近最优调度时间长度,并且能够处理更大规模的应用负载,实用性更强。

By applying the approaches for the overlap execution of computation and communication and multiple parallel distribution tasks, this paper presented a multi-round algorithm for scheduling divisible workloads on the heterogeneous cluster computing systems that processors had different computing speeds and communication capabilities and memory capacities. The experiment results on the cluster of personal computers show that the presented algorithm has practicality, and it not only obtains the asymptotically optimal scheduling time length as the uniform multi-round algorithm, hut also can process more largescale workloads.