云环境下虚拟机部署算法的研究

发布时间：2018-03-18 08:41

本文选题：云计算　切入点：虚拟机部署　出处：《天津工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：在云数据中心虚拟机的部署策略已成为一个重要的研究课题,也是当前研究领域的一个热点和难点问题。在云数据中心设有大量拥有的计算资源量是各不相同的物理服务器,如果不对这些服务器和携带任务的虚拟机进行分析直接将虚拟机随机部署到服务器上会导致云计算系统的负载平衡性能差、虚拟机执行效率低,而且会大量地浪费物理资源。本文首次提出了一种基于贝叶斯分类并结合首次自适应的思想进行虚拟机部署的策略。该策略算法通过对虚拟机需求以及物理服务器的剩余资源进行分析,利用贝叶斯公式及标准化的阈值找出能达到长期负载平衡适合部署虚拟机的物理服务器,将虚拟机的需求量进行降序排序后以首次适应算法部署到物理服务器上不仅降低了虚拟机部署的失败次数,也减少启动了不必要开启的物理服务器从而降低了整个系统的能耗。仿真实验结果表明,新的任务部署策略不仅更好地实现了长期负载平衡,缩短了任务的完成时间,而且在降低能耗和任务部署失败率方面都有明显的效果。本文首次提出了利用朴素贝叶斯公式将虚拟机进行进一步地分类工作后,以先来先服务的方式进行虚拟机的部署。首先通过对虚拟机需求以及物理服务器的剩余资源进行分析,利用贝叶斯公式筛选出具有高效性能的服务器集群,再将这些服务器集群作为统计朴素贝叶斯所需概率的资源池,当后续再对虚拟机进行分配时,则根据朴素贝叶斯的分类策略就可以快速的找出可部署虚拟机的服务器集,将虚拟机以先来先服务的算法部署到分类后的服务器上,既可以达到负载平衡的效果也能使得服务器集里的服务器负载均衡,同时也提高了整个系统的执行效率。仿真实验结果表明,该任务部署策略不仅很好地实现了系统的长期负载平衡,缩短了任务的完成时间,而且在降低任务.部署失败率方面也有明显的效果。
[Abstract]:The deployment strategy of virtual machine in cloud data center has become an important research topic, and it is also a hot and difficult problem in current research field. If you don't analyze these servers and the virtual machines that carry tasks, deploying the virtual machines directly to the server at random will lead to poor load balancing performance of cloud computing systems, and the execution efficiency of virtual machines will be low. This paper presents a strategy of virtual machine deployment based on Bayesian classification and first adaptive thinking. The strategy algorithm is based on the requirements of virtual machines and physical services. To analyze the remaining resources of the. Using Bayesian formulas and standardized thresholds to find physical servers that can achieve long-term load balancing suitable for deploying virtual machines, Ordering the requirements of virtual machines in descending order and deploying the first adaptive algorithm to the physical server not only reduces the number of virtual machine deployment failures, The simulation results show that the new task deployment strategy not only realizes the long-term load balance better, but also shortens the completion time of the task. Moreover, it has obvious effect in reducing energy consumption and mission deployment failure rate. In this paper, we first put forward using naive Bayes formula to classify virtual machines further. First of all, by analyzing the requirement of virtual machine and the remaining resources of physical server, the server cluster with high efficiency is selected by using Bayesian formula. Then these server clusters are used as resource pools to calculate the probability of naive Bayes. When the virtual machines are allocated later, the server set of deployable virtual machines can be quickly found according to the classification strategy of naive Bayes. Deploying the virtual machine as a first-come, first-served algorithm to the classified servers can not only achieve the effect of load balancing, but also make the server load balance in the server set. At the same time, the efficiency of the whole system is improved. The simulation results show that the task deployment strategy not only realizes the long-term load balance of the system, but also shortens the completion time of the task. It also has a significant effect on reducing mission-deployment failure rates.
【学位授予单位】：天津工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP302

【参考文献】