云计算环境下能源感知任务调度策略研究

【作者】 李慧；

【导师】 雷丽晖；

【作者基本信息】 陕西师范大学 ， 计算机技术（专业学位）， 2015， 硕士

【摘要】 云计算是一种基于互联网的商业计算模式,具有动态伸缩性的特点。云计算环境由一系列能够动态伸缩的资源构成,云计算服务提供商通过虚拟化技术将这些资源提供给云计算用户。用户可以按需租赁云计算资源,这样不仅可以减少自身终端处理负担,而且可以拥有云计算的强大的计算能力。当人们利用云计算环境中的资源处理海量任务时,云计算环境下合理的任务调度策略成为提高任务执行效率、充分利用网络资源的有效手段。因此,对云计算环境下的任务调度策略的研究有着重要的意义。本文对云计算以及云计算环境下的任务调度进行了深入研究,针对现有云计算环境下的任务调度策略缺乏考虑用户任务偏好,从而导致虚拟机资源利用不充分、用户对服务质量满意度不高等问题,提出了云计算环境下能源感知任务调度策略。本文的主要的研究工作如下：(1)对云计算环境下的任务调度环境进行系统建模,即分别对用户、任务、及资源的模型进行了描述。(2)定义了任务偏好指数计算公式,对用户任务信息进行计算得到用户任务偏好指数,具体方法为：首先计算出用户任务的平均偏好值；然后求得用户任务信息与平均偏好值之间的偏差率；最后,将偏差率存储为用户任务的偏好指数。(3)定义了性能评分计算公式,对虚拟机资源性能信息进行计算,求得资源性能评分,具体方法为：首先计算出虚拟机资源性能的平均性能值；然后求得虚拟机资源性能与平均性能值之间的偏差率；最后将偏差率存储为虚拟机资源性能评分。(4)依据马氏距离公式计算任务偏好指数和资源性能评分之间的马氏距离,同时计算能源消耗量,根据任务的类型计算任务调度指标值。在任务调度过程中,为任务选择任务调度指标值最小的虚拟机资源进行映射,并且用马氏距离的值作为用户任务满意度的衡量指标。(5)最后,通过扩展CloudSim云计算仿真平台的Cloudlet类、Vm类以及DataCenterBroker类,实现了本文提出的能源感知任务调度策略；通过对CloudSim平台进行了重编译生成,并在生成的扩展平台上实现了能源感知任务调度策略的仿真算法,对文中提出的策略进行了模拟验证和对比分析。实验结果显示该算法能够有效提高任务执行效率和用户任务满意度,并且降低环境中的能源消耗。更多还原

【Abstract】 Cloud computing is a computing model based on commercial Internet, with dynamic scalability characteristics. Cloud computing environment consists of a series of dynamically scalable resources, the cloud computing service provider will provide these resources to cloud computing users with virtualization technology. Users can lease cloud computing resources on demand, it not only can reduce the processing load on the terminal itself, but also can have a cloud computing power. When people deal with the massive task in cloud computing environment, a reasonable scheduling strategy in cloud computing environments becomes an efficient and effective means to improve the task execution and to fully utilize network resources. Therefore, the research of task scheduling strategy in cloud computing environment has important significance.In this paper, after in-depth research of cloud computing and cloud computing task scheduling strategy in cloud computing environments, in connection with the lack of consideration of user preferences in the existing task scheduling strategy in cloud computing environment, so that the virtual machine resource utilization is not sufficient, user satisfaction of the quality of service is a problem, this paper proposed an energy-awared scheduling strategy in cloud computing environment.The main research work is as follows:(1) Models the task scheduling environment in cloud computing environment for system, which consists of users, tasks, and resources.(2) Defines the task preference index which is calculated with the user task information. The specific method is:First calculate the average value of the user preference of the task; then obtained the deviation rate of task information between the user preference and the average value of preference; Finally, the error rate is stored as user task preference index.(3)Defines the virtual machine resource performance score which is calculated by the performance parameters of the resources properties.The specific method is:First calculate the average performance value of the virtual machine resource properties; then obtain the deviation rate of a virtual machine resource bwteen performance parameters and average performance value; and finally the deviation rate is stored as the virtual machine resource performance score.(4) Firstly,calculate the Mahalanobis distance between task preferences and resource performance index score with the Mahalanobis distance formula.And then calculates energy consumption.Finally,calculate index value depending on the type of task. In the scheduling process, select the virtual machine of the minimum index value resources for the tasks, and use the value of the Mahalanobis distance as a measure of user satisfaction task.(5) Finally, by extending Cloudlet class, Vm class and DataCenterBroker class in CloudSim so that the proposed energy-aware scheduling strategy can be implement; CloudSim platform is compiled.Experimental results show that scheduling strategy proposed in this paper can improve the efficiency of task execution and user tasks satisfaction, and reduce energy consumption in the environment.更多还原