基于二次聚类的多目标混合云任务调度算法

发布时间：2019-05-28 01:32

【摘要】：针对混合云环境包含大量异构云计算节点的情况,提出二次聚类方法,依据资源的综合特性,将异构资源进行分簇,将任务分发到合适的资源聚类,缩小任务搜索空间.在此基础上,结合私有云的安全可靠性、公有云的可扩展性以及用户需求的多样性,提出混合云环境下多目标优化的任务调度算法.该算法首先在私有云优先调度截止时间短的任务,对于每个聚类,将任务分配给完成时间最接近于其结束时间的资源,以完成更多的任务;将溢出的高负载任务转移到公有云聚类执行,结合任务的计算成本、通信开销和截止时间的约束,选择费用最低的资源.实验结果表明,与传统无聚类的算法相比,该算法降低了执行费用,同时提高了资源利用率和用户满意度.
[Abstract]:In view of the fact that the hybrid cloud environment contains a large number of heterogeneous cloud computing nodes, a secondary clustering method is proposed. According to the comprehensive characteristics of the resources, the heterogeneous resources are grouped, the tasks are distributed to the appropriate resource clustering, and the task search space is reduced. On this basis, combined with the security and reliability of private cloud, the scalability of public cloud and the diversity of user requirements, a multi-objective optimization task scheduling algorithm in hybrid cloud environment is proposed. First of all, the algorithm gives priority to scheduling tasks with short deadline in the private cloud. For each clustering, the task is allocated to the resources whose completion time is closest to its end time in order to complete more tasks. The overflow high load task is transferred to the public cloud clustering to execute, and the resources with the lowest cost are selected according to the constraints of task calculation cost, communication overhead and deadline. The experimental results show that compared with the traditional clustering-free algorithm, this algorithm reduces the execution cost and improves the resource utilization and user satisfaction.
【作者单位】：北京交通大学交通数据分析与挖掘北京市重点实验室;
【基金】：国家自然科学基金资助项目(61300176) 中央高校基本科研基金资助项目(2016JBM019)
【分类号】：TP301.6

【相似文献】