移动云计算的系统管理与优化方法研究

发布时间：2018-03-26 19:38

本文选题：移动云计算　切入点：资源聚类　出处：《吉林大学》2017年硕士论文

【摘要】：随着移动互联网和云计算的快速发展,移动云计算成为了最新的研究热点。智能移动设备在各项性能和处理能力上的大幅提高,以及分布式处理技术如Hadoop和Map Reduce在移动设备上的应用,使移动设备逐渐开始作为移动云资源池中的服务器节点参与高性能计算并处理大量数据。而一方面移动设备本身具有移动性,移动网络覆盖也存在间隙,不同的设备在处理任务时表现出不同的稳定性与处理能力;另一方面不同的任务在各项资源上有着不同的需求,为移动设备资源的统一管理和分配带来了挑战。为此,本文对移动云计算的资源管理与任务分配问题进行了较深入的研究。本文的研究内容以及创新点如下:(1)选取一种基于代理服务器的移动云计算系统架构,利用代理连接移动设备和远端云。并在代理服务器中设置了监控器、资源管理器、任务管理器和调度器等功能模块来进行可靠的资源管理以及进行高效的任务分配工作。(2)提出了一种基于改进的K-means算法的移动资源聚类方法PKMC_RM。综合考虑移动资源实际的移动特性、可靠性、资源受限等问题,将移动设备资源按其移动率和系统出错率进行聚类,估计每个移动设备簇的处理能力,每个移动设备簇根据自身处理能力执行有不同资源需求的任务。给出了PKMC_RM的详细设计和实现,包括输入参数的选定和标准化、距离的计算,和在初始聚类中心的选择、新聚类中心的查找以及判断是否继续查找规则上的改进,实验证明PKMC_RM能够达到较好的聚类效果从而满足资源聚类管理的可靠性要求。(3)提出一种基于蚁群算法的多目标任务分配方法MOAC_TD,分两个阶段完成任务的最优分配。给出了MOAC_TD详细的设计和实现,包括任务的聚类过程、选取负载均衡和最低电能消耗作为两个优化目标、适应度函数的设计、信息素更新规则的改进和状态转移概率的设计、以及在每轮迭代中利用Pareto支配关系寻找最优解等。实验表明MOAC_TD能够在满足任务对资源的需求基础上,较好的保持系统负载的均衡并且有较好的节能效果。
[Abstract]:With the rapid development of mobile Internet and cloud computing, mobile cloud computing has become the latest research hotspot. And the application of distributed processing technologies such as Hadoop and Map Reduce in mobile devices, Mobile devices gradually begin to participate as server nodes in mobile cloud resource pool in high performance computing and processing a lot of data. On the one hand, mobile devices have mobility, and there are gaps in mobile network coverage. Different devices have different stability and processing ability when dealing with tasks, on the other hand, different tasks have different demands on each resource, which brings challenges to the unified management and allocation of mobile device resources. In this paper, the resource management and task allocation of mobile cloud computing are deeply studied. The research content and innovation of this paper are as follows: 1) choose a mobile cloud computing system architecture based on proxy server. Connect mobile devices and remote clouds with agents, and set up monitors, resource managers in proxy servers, This paper presents a mobile resource clustering method based on improved K-means algorithm. Based on the problems of reliability and resource limitation, the mobile device resources are clustered according to their mobility rate and system error rate, and the processing power of each mobile device cluster is estimated. Each cluster of mobile devices performs tasks with different resource requirements according to its processing capacity. The detailed design and implementation of PKMC_RM are presented, including the selection and standardization of input parameters, the calculation of distances, and the selection of initial clustering centers. The search of the new clustering center and the improvement of judging whether to continue to look up the rules, Experimental results show that PKMC_RM can achieve better clustering effect and meet the reliability requirements of resource clustering management.) A multi-objective task allocation method based on ant colony algorithm, MOACT _ D, is proposed. The optimal assignment of tasks is completed in two stages. The detailed design and implementation of MOAC_TD are given. Including task clustering process, load balancing and minimum power consumption as two optimization objectives, fitness function design, pheromone update rule improvement and state transition probability design. The experiment shows that MOAC_TD can keep the balance of system load and save energy on the basis of satisfying the demand of resource.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;TP18

【相似文献】