面向云工作流的两阶段资源调度方法

发布时间：2018-02-16 00:26

本文关键词： 云计算工作流粒子群算法负载感知　出处：《华南理工大学学报(自然科学版)》2017年01期 　论文类型：期刊论文

【摘要】：为提高云计算系统的资源利用率,优化系统性能,同时兼顾用户的服务质量(Qo S)需求约束,文中结合云计算和工作流建立了云工作流系统,给出了具有两个调度阶段的系统资源调度模型.在第1阶段中,考虑了Qo S的时间及价格约束、工作流内各个任务之间的依赖关系以及各个任务所产生的中间数据的处理,提出了改进的粒子群优化(MPSO)算法,并利用Pareto获得最优解,以提高调度效率;在第2阶段中,考虑了资源在主机上的分配情况,提出了具有负载感知的调度策略,根据系统的负载情况进行资源调度,以提高系统的资源利用率.实验结果表明:在云工作流系统的资源优化调度中,与经典的异构最早完成时间算法、单目标优化的遗传算法相比,MPSO算法的任务执行速度更快、资源利用率更高,能满足用户的Qo S需求;具有负载感知的调度策略能更有效地根据负载情况进行调度,提高任务执行的效率和资源利用率.
[Abstract]:In order to improve the resource utilization of cloud computing system, optimize the system performance and take into account the quality of service (QoS) requirement constraints of users, a cloud workflow system is established based on cloud computing and workflow. The system resource scheduling model with two scheduling stages is given. In the first stage, the time and price constraints of QoS, the dependencies between tasks in workflow and the processing of intermediate data generated by each task are considered. An improved particle swarm optimization (PSO) algorithm is proposed, and the optimal solution is obtained by using Pareto to improve the scheduling efficiency. In the second stage, considering the allocation of resources on the host, a load-aware scheduling strategy is proposed. Resource scheduling is carried out according to the load of the system in order to improve the resource utilization of the system. The experimental results show that in the resource optimization scheduling of the cloud workflow system, it is the earliest completion time algorithm with the classic isomerism. Compared with MPSO, the single objective optimization genetic algorithm has higher task execution speed and higher resource utilization, which can meet the needs of users' QoS, and the load-aware scheduling strategy can be more effectively scheduled according to the load situation. Improve task execution efficiency and resource utilization.
【作者单位】：东北大学信息科学与工程学院;东北大学秦皇岛分校控制工程学院;
【基金】：国家自然科学基金资助项目(61300194)~~
【分类号】：TP18

【相似文献】