面向虚拟机动态部署的服务性能预测和部署方案研究

发布时间：2018-09-19 07:13

【摘要】：云计算是以虚拟化技术为基础,将物理资源虚拟成一个动态的拟资源池,以服务的形式提供给用户使用。虚拟化技术能够有效地增强服务的可靠性,并能显著简化IT基础设施、优化资源以及降低能耗。但是,虚拟化技术引起的物理架构变化和虚拟机上服务性能变化,给保证云服务的质量工作带来了很大的挑战。因此,如何根据虚拟机上服务的性能变化趋势和用户的访问情况,合理地为各服务分配所需的虚拟机实例数,并为其动态生成虚拟机部署方案,以满足服务的性能需求是云计算中资源调度及部署研究的重要课题之一为实现云环境下各服务对虚拟机资源需求的有效管理,本文对云环境下面向虚拟机动态部署的服务性能预测和部署方案的生成方法这一课题进行了研究,根据云计算中各虚拟机上运行的服务性能变化趋势,动态地为其分配虚拟机实例并生成合理的动态部署方案,达到高资源利用率、低服务成本、保证用户体验和服务质量的目标。本文的研究工作主要有：第一,本文研究了基于服务性能预测的虚拟机动态部署过程。分析了传统虚拟机部署研究的不足并进行改进,给出该过程中涉及的重要环节,如服务性能变化趋势的预测和虚拟机动态部署方案的生成。第二,提出了面向虚拟机动态部署的服务性能变化趋势的预测算法。主要研究内容有确定服务性能变化的影响因子,以及利用基于MGM(1,N)-BP神经网络的预测模型进行预测。第三,提出了基于改进的蚁群算法的虚拟机动态部署方案的生成方法。对虚拟机动态部署方案生成问题的优化目标进行了分析,建立综合优化模型,然后通过改进智能优化算法生成合理的部署方案。实验表明,本文提出的云环境下面向虚拟机动态部署的服务性能预测算法和部署方案生成方法是可行的,并且具有一定的准确性和有效性。
[Abstract]:Cloud computing is based on virtualization technology, virtual physical resources into a dynamic quasi-resource pool, in the form of services for users to use. Virtualization can effectively enhance the reliability of services, significantly simplify the IT infrastructure, optimize resources, and reduce energy consumption. However, the changes of physical architecture and service performance on virtual machines caused by virtualization technology bring great challenges to ensure the quality of cloud services. Therefore, how to reasonably assign the number of virtual machine instances for each service according to the changing trend of service performance and user access, and dynamically generate the virtual machine deployment scheme for each service. To meet the performance requirements of services is one of the most important research topics of resource scheduling and deployment in cloud computing. In this paper, the prediction of service performance for dynamic deployment of virtual machines and the generation of deployment schemes in cloud environment are studied. According to the trend of service performance in cloud computing, The virtual machine instance is dynamically assigned and a reasonable dynamic deployment scheme is generated to achieve the goal of high resource utilization, low service cost, and guaranteeing user experience and quality of service. The main work of this paper is as follows: first, the dynamic deployment process of virtual machine based on service performance prediction is studied. The shortcomings of the traditional virtual machine deployment research are analyzed and improved, and the important links involved in the process are given, such as the prediction of the trend of service performance change and the generation of the virtual machine dynamic deployment scheme. Secondly, a dynamic deployment oriented service performance prediction algorithm is proposed. The main content of this paper is to determine the influence factors of service performance change, and to use the prediction model based on MGM (1N) -BP neural network to predict the change of service performance. Thirdly, a dynamic deployment scheme of virtual machine based on improved ant colony algorithm is proposed. In this paper, the optimization goal of dynamic deployment scheme generation of virtual machine is analyzed, and a comprehensive optimization model is established, and then a reasonable deployment scheme is generated by an improved intelligent optimization algorithm. Experiments show that the proposed service performance prediction algorithm and deployment scheme generation method for dynamic deployment of virtual machines in the cloud environment are feasible and have certain accuracy and effectiveness.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP302

【相似文献】