云计算环境下面向任务分类的个性虚拟化策略

发布时间：2018-03-08 00:22

本文选题：云计算　切入点：任务分类　出处：《辽宁大学》2012年硕士论文　论文类型：学位论文

【摘要】：提高资源利用率是云计算的一个重要研究方向。目前，主要是通过人工智能方法对任务的资源需求做出预测，并根据服务器资源的使用情况选择最优的任务分配方案，来提高资源的利用率。这些方法不仅和平台密切相关，而且在任务数量巨大、以虚拟机为资源调度单位的云计算环境下，这些方法需要耗费大量的资源来采集、存储、加工数据，整合和迁移虚拟机。本文提出的云计算环境下面向任务分类的个性虚拟化策略，通过综合分析云计算中的任务，并结合已有的通信协议、程序度量、资源预测等分析方法估测任务对处理器、网络带宽、磁盘等资源的需求特性，然后根据任务的资源需求特性将云计算中的任务分为计算型、通信型、磁盘型、计算通信型、计算磁盘型、通信磁盘型、强标准型、弱标准型等多种类型；对任务分类以后，再结合现有的虚拟化技术，按照特定的资源配置比例给虚拟机分配处理器、网络带宽、磁盘等资源，从而虚拟出和任务类型相对应的个性化虚拟机；然后，提出了针对任务分类和个性化虚拟机的个性虚拟化策略，该策略按照一定的约束条件，根据需要构建新的虚拟机或者回收已有的虚拟机资源，并在需要启动物理或者关闭物理机时启动和关闭物理机。和传统的基于资源预测的负载均衡方法相比，本文提出的策略可以避免一些资源不能满足需求而其它资源却处于空闲状态的情况出现，并提高任务的执行效率和资源的整体利用率。最后，在不同的硬件配置环境下对任务分类系统的平台无关性进行验证，，并在模拟云计算的环境下，对比了在采用个性虚拟化策略和没有采用该策略的条件下资源的使用情况。结果证明，本文提出的策略在云计算环境下，能够有效性的提高资源利用率。
[Abstract]:Improving resource utilization is an important research direction of cloud computing. At present, it mainly uses artificial intelligence method to forecast the resource demand of the task, and selects the optimal task allocation scheme according to the use of server resources. These methods are not only closely related to the platform, but also need a large amount of resources to collect and store in the cloud computing environment where the virtual machine is the resource scheduling unit. Process data, integrate and migrate virtual machines. This paper proposes a personalized virtualization strategy for task classification in cloud computing environment. By synthetically analyzing the tasks in cloud computing, and combining the existing communication protocols, program metrics, resource prediction and other analysis methods to estimate the task to processor. Network bandwidth, disk and other resource requirements, and then according to the resource requirements of the task, the tasks in cloud computing are divided into computational type, communication type, disk type, computational communication type, computing disk type, communication disk type, strong standard type. After classifying tasks and combining existing virtualization technologies, the virtual machine is allocated resources such as processor, network bandwidth, disk and so on according to specific resource allocation ratio. Then, a personalized virtual machine corresponding to the task type is proposed. Then, a personalized virtualization strategy for task classification and personalized virtual machine is proposed, which is based on certain constraints. Build new virtual machines as needed or recycle existing virtual machine resources, and start and shut down physical machines when they need to be started or shut down. The strategy proposed in this paper can avoid the situation that some resources can not meet the requirements while others are idle and improve the efficiency of task execution and the overall utilization of resources. Finally, the platform independence of task classification system is verified in different hardware configuration environment, and in the environment of simulating cloud computing, The results show that the proposed strategy can effectively improve resource utilization in cloud computing environment.
【学位授予单位】：辽宁大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP3

【引证文献】