异构云平台中基于多层架构的动态循环调度方案

针对现有分布式循环自调度方案在异构云平台中存在负载不平衡等问题,提出一种基于多层架构的分层分布式动态循环调度方案。首先,通过HPLS算法来评估计算环境中各Worker节点的计算速度。然后,在传统自调度方案中融入节点计算速度,构建一种能够处理异构环境的调度方案,提高负载平衡能力。最后,将计算系统构建成一个由SuperMaster,Master和Worker节点组成的多层架构,利用层次化方法来解决传统Master-Worker架构中单个Master节点的瓶颈问题,用来提高任务分配效率。仿真实验结果表明,提出的方案能够有效提高云平台的计算效率。

A Dynamic Loop Scheduling Scheme in Heterogeneous Cloud Based on Multi-layer Architecture

For the issues that the load imbalance of existing distributed loop self-scheduling scheme in heterogeneous cloud, this paper proposed a hierarchical distributed dynamic loop scheduling scheme. Firstly, it used the HPLS algorithm to evaluate the computational speed of each Worker node in the computing environment. Then, it integrated the node computing speed with the traditional self-scheduling scheme, so as to deal with the heterogeneous environment and improve the load balancing capability. Finally, it constructed the computing system as a multi-layer architecture composed of SuperMaster, Master and Worker nodes, which can solve the single master node bottleneck problem of traditional Master-Worker architecture, and to improve the task allocation efficiency.The simulation results show that the scheme can effectively improve the computational efficiency of the cloud platform.