移动协作微云计算中的资源分配研究

发布时间：2018-01-21 08:07

本文关键词： 移动云协作微云计算资源分配通信资源分配　出处：《东南大学》2016年硕士论文　论文类型：学位论文

【摘要】：智能移动设备逐渐成为人们生活中不可或缺的一部分。但是,它始终无法克服便携性与自身资源有限之间的矛盾,直到移动云计算被提出。在移动云计算模型中,移动应用被转移至云资源处理,从而达到节约移动终端自身资源的目的。目前移动云计算中可利用云资源包括：远程公共云、临近微云。相较于远程云计算中心的资源,临近微云具有通信时延短的优点,但是也具有单点计算资源有限的缺点。针对临近微云单点计算资源有限这一缺陷,本文提出移动协作微云计算(M3C, Mobile Cooperative Cloudlet Computing)系统,在该系统中,来自移动用户终端的应用被表示为由子任务以及子任务之间的数据约束组成的工作流,通过将子任务被分配到系统中的多个接入点分别处理,提高工作流处理的并行度,缩短用户等待处理结果的时间。本论文研究的主要内容是M3C系统的计算资源与通信资源分配方案。首先,本文将已有的粒子群搜索(PSO, Particle Swarm Optimization)算法、异构最短结束时间(HEFT, Heterogeneous Earliest- Finish-Time)算法、部分关键路径(PCP, Partial Critical Path)算法应用于M3C系统中。其中,PSO算法属于随机搜索算法,该算法在迭代次数较多时可以达到较好的结果,但是计算量大。HEFT算法与PCP算法均属于列表启发式算法,分配过程包括两部分：计算列表、按列表次序分配。不同的列表启发式算法区别在于列表排序策略和分配策略。HEFT中按照任务优先级进行排序,再将任务分配到能够最早完成该任务的接入点；PCP算法在任务优先级的基础上,将任务划归到若干个集合中,以集合(PCP)为单位进行分配,同一个集合中的任务被分配到同一个接入点。其次,本文以计算资源分配为基础,获得系统无线链路集合,存在无线通信需求的接入点之间的距离需小于距离阈值,无线链路集合对应了一组接入点间距离约束条件。本文分别利用穷举和动态规划方法对接入点进行部署,接入点位置不仅满足距离约束条件且覆盖面积最大化。在进行信道分配时,考虑链路间的相互干扰,避免同一条信道被相互干扰的链路同时占用,通过多次循环实现均匀分配信道。最后,设计一系列仿真实验用于评估系统性能。每次实验包括四步：计算资源分配、接入点部署、信道资源分配以及运行模拟。仿真实验的内容包括：M3C系统与ICloudlet系统对比仿真,以证明接入点之间的协作能够缩短应用工作流处理时间；控制变量法分析M3C系统中的接入点个数、服务器切片数、计算资源量、信道假设以及通信资源量对系统性能的影响,仿真结果表明,系统性能不仅取决于计算资源与通信资源的分配情况,也和计算资源与通信资源比例有关,合理设计计算与通信资源比例,才能有效提高系统性能。
[Abstract]:Intelligent mobile devices have gradually become an indispensable part of people's lives. However, it can not overcome the contradiction between portability and its limited resources. In the mobile cloud computing model, mobile applications are transferred to cloud resource processing. At present, the cloud resources can be used in mobile cloud computing, including: remote public cloud, near micro-cloud. Compared with the resources of remote cloud computing center. Proximity microcloud has the advantage of short communication delay, but it also has the disadvantage of limited single-point computing resources. Aiming at the limitation of single-point computing resources, mobile cooperative micro-cloud computing is proposed in this paper. The Mobile Cooperative Cloudlet computing system. An application from a mobile user terminal is represented as a workflow composed of sub-tasks and data constraints between sub-tasks, which are processed separately by assigning sub-tasks to a plurality of access points in the system. Improve the parallelism of workflow processing and shorten the time for users to wait for processing results. The main content of this paper is the allocation of computing and communication resources in M3C system. First of all. In this paper, the existing particle swarm optimization (PSO) algorithm, Particle Swarm optimization algorithm, is used to calculate the minimum end time of isomerism. Heterogeneous Earliest-Finish-Timealgorithm, some critical paths. Partial Critical path algorithm is applied to M3C system. The algorithm can get better results when the number of iterations is more, but the algorithm of. Heft and PCP both belong to list heuristic algorithm, and the assignment process includes two parts: computing list. The difference between different list heuristic algorithms is that the list sorting strategy and the assignment strategy .HEFT sort according to the priority of the task. The task is then assigned to the access point that can finish the task as early as possible; On the basis of the priority of tasks, the PCP algorithm divides the tasks into several sets and assigns the tasks in the same set to the same access point. Secondly, the tasks in the same set are assigned to the same access point. On the basis of computing resource allocation, the system wireless link set is obtained. The distance between access points with wireless communication needs to be less than the distance threshold. The wireless link set corresponds to a set of distance constraints between access points. In this paper, we use exhaustive and dynamic programming methods to deploy access points. The location of the access point not only satisfies the distance constraint condition but also maximizes the coverage area. In the channel allocation, the interference between links is considered to avoid the same channel being occupied by the interference link simultaneously. Finally, a series of simulation experiments are designed to evaluate the performance of the system. Each experiment consists of four steps: computing resource allocation, access point deployment. Simulation of channel resource allocation and operation. The simulation results show that the collaboration between access points can shorten the processing time of application workflow. The control variable method is used to analyze the influence of the number of access points, the number of server slices, the amount of computing resources, the channel assumption and the amount of communication resources on the performance of M3C system. The system performance depends not only on the allocation of computing resources and communication resources, but also on the ratio of computing resources to communication resources.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN929.5;TP3

【参考文献】