IaaS下虚拟机弹性服务与部署的研究

发布时间：2018-11-20 04:08

【摘要】：当前,云计算技术发展迅速,处于技术攻坚时期,其研究与应用将对未来的信息技术产业产生深远的影响。基础设施即服务(Infrastructure as a Service, IaaS)作为一种基本的资源供应服务,是云计算技术发展的基础。研究发现,IaaS云服务存在服务效率低下、用户体验较差、数据中心资源利用率低下等问题,而弹性云的优越性也渐渐明显。因此,开展IaaS弹性云服务的研究具有科学意义和研究价值。本文分析了当前弹性云的技术现状和发展趋势,发现其实现机制的现存问题是造成云服务效率低下、响应延迟的主要因素,在供应虚拟机的IaaS云计算环境下,虚拟机是物理资源的直接占用者,合理的部署是提高资源利用率的途径。为了提高弹性云服务效率和资源利用率等问题,研究从弹性云实现机制入手,探讨了IaaS下基于预测的虚拟机弹性服务机制与虚拟机部署的实现方法,主要开展了以下方面的研究工作： (1)分析弹性云服务效率问题,提出了一种基于ARIMA模型的负载预测算法。算法首先借助ARIMA模型在时间序列分析方面的优势和季节指数量化序列的特点,对负载进行预测和调整,能较精确地预测出云计算环境下基于Web的负载数据；然后根据负载与虚拟机配置的关系预算出所需供应的虚拟机的配置和数量,组成基于预测的虚拟机资源预留决策,为实现弹性云提供一种可行的机制。为验证该算法的有效性,研究中引入了负载压力测试工具Loadrunner以及第二次世界杯足球赛Web站点的部分日志数据。LoadRunner用于模拟产生仿真负载,对算法得到的预留虚拟机加载特定的负载,测试所预留的虚拟机数量是否合理。测试数据结果表明,基于本文算法生成的弹性机制是一种有效、可行的方案,适用于弹性云的虚拟机预留决策。 (2) IaaS云计算环境下虚拟机的部署是有关资源管理、服务质量等方面的综合性问题,已有的研究已经证明这是一个多目标的NP-hard问题,本文研究发现,虚拟机的部署关键在于如何设定部署算法实现的多个目标,将实际云计算环境下的需求转化为算法可识别的信息。为此,本文以弹性云框架为研究对象,综合考虑虚拟机部署对整个云计算框架的影响,针对现有研究存在的不足以及新形势下虚拟机部署面临的新问题,利用遗传算法解决多目标组合优化问题的优点,提出了基于改进遗传算法的虚拟机部署方案。该方案以减少物理计算机开销、提高资源利用率、满足用户SLA为优化目标,改进遗传算法,使之符合解决该部署问题的要求。实验结果分析表明,本文提出的基于改进遗传算法的虚拟机部署方案可以有效地提高资源利用率、降低物理节点开销、降低服务代价,同时算法适应静态和动态部署要求,为虚拟机到物理节点的映射提供了较好的解决方案。本文的研究为IaaS下虚拟机弹性服务机制与部署提供了新的方法和研究思路,研究成果具有科学意义和研究价值。
[Abstract]:At present, the development of cloud computing technology is fast, and the research and application of cloud computing technology will have a far-reaching impact on the future information technology industry. Infrastructure as a service (IaaS) as a basic resource-supply service is the foundation of the development of cloud computing technology. The research has found that the IaaS cloud service has the problems of low service efficiency, poor user experience, low utilization rate of data center resources and the like, and the advantages of the elastic cloud are becoming more and more obvious. Therefore, the research of IaaS elastic cloud service is of scientific significance and research value. In this paper, the present situation and development trend of the present elastic cloud are analyzed, and the existing problems of the implementation mechanism are found to be the main factors that cause the low efficiency of the cloud service and the response delay. In the IaaS cloud computing environment of the supply virtual machine, the virtual machine is the direct account of the physical resources. The rational deployment of the user is to improve the utilization of the resources In order to improve the efficiency of the elastic cloud service and the utilization of resources, the paper studies the realization method of the virtual machine elastic service mechanism and the virtual machine deployment based on the forecast under the IaaS, and mainly carries out the following aspects: Work: (1) To analyze the problem of elastic cloud service efficiency, and put forward a negative analysis method based on ARIMA model The algorithm is based on the advantage of ARIMA model in the time series analysis and the characteristics of the season finger quantization sequence, and the load is predicted and adjusted, and the Web based on the Web can be predicted more accurately. and then, according to the relationship between the load and the virtual machine configuration, the configuration and the number of the virtual machines to be supplied are calculated, feasible mechanism. In order to verify the effectiveness of the algorithm, the load pressure test tool Loadrunner and the second World Cup soccer match Web site are introduced. Split log data. LoadRunner is used to simulate the generation of simulated load, load specific load on the reserved virtual machine obtained by the algorithm, and test the number of virtual machines reserved The test results show that the elastic mechanism based on this algorithm is an effective and feasible solution, which is suitable for the virtual cloud. Machine reservation decision. (2) The deployment of virtual machine under the IaaS cloud computing environment is a comprehensive problem in the aspects of resource management and service quality. The existing research has proved that this is a multi-objective NP-hard question In this paper, it is found that the key to the deployment of the virtual machine is how to set the target of the deployment algorithm, and transform the demand of the real cloud computing environment into the actual cloud computing environment. In this paper, the paper takes the elastic cloud framework as the research object, and considers the influence of the virtual machine deployment on the whole cloud computing framework, aiming at the defects existing in the existing research and the new problems facing the deployment of the virtual machine under the new situation, and solving the multi-target group by using the genetic algorithm. Based on the improved genetic algorithm, the advantages of the optimization problem are put forward. The scheme is to reduce the physical computer cost, improve the resource utilization rate, meet the user's SLA as the optimization target, and improve the genetic algorithm to solve the problem. The results of the experiment show that the proposed virtual machine deployment scheme based on the improved genetic algorithm can effectively improve the resource utilization rate, reduce the physical node overhead, reduce the service cost, and also adapt the algorithm to the static State and dynamic deployment requirements, providing mapping for virtual machines to physical nodes The research of this paper provides a new method and a new approach to the mechanism and deployment of virtual machine elastic service under IaaS, and the research results
【学位授予单位】：广西大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP302

【相似文献】