可靠虚拟数据设计及研究

发布时间：2019-05-23 11:38

【摘要】：面向云计算的虚拟数据中心(Virtual data center: VDC)以其低廉的价格和非常有吸引力的按需付费模式受到了广泛的关注。越来越多的企业从部署专用服务器的传统运营模式转向租用云数据中心的资源来满足企业日常的IT应用需求，，以此摆脱维护和管理基础设施网络的高昂开支。通过虚拟化技术，这些租户可以高效地共享同一个基础设施的资源，获得了更经济的基础设施即服务(InfrastructureAsServices: IaaS)。与传统的数据中心模式相比，云计算环境下资源虚拟化技术在提高资源的利用率以及降低能耗等方面都展现出了巨大的优势。但是其高度整合的资源组织和管理方式也带来了潜在的危机。全球每年都有成百上千的数据中心发生服务中断事件，导致重大损失。同时，为降低成本，云数据中心通常采用商用而非电信级的底层硬件，其可靠性本身就存在问题，虚拟数据中心的高可靠性已成为目前学术和产业界都亟待解决的问题。本文的研究了以下三个问题：(1)当链路发生故障时虚拟数据中心的冗余设计，目的是保证任意单链路失效场景下虚拟数据中心的可用性；(2)当虚拟机发生故障（虚拟机本身故障或承载虚拟机的物理服务器故障）时虚拟数据中心的冗余设计，目的是实现单虚拟机失效时虚拟数据中心的可用性；(3)利用开源软件Pacemaker完成高可靠性虚拟数据中心平台的搭建，通过不同的冗余设置，实现应用级和虚拟机级的高可靠性。针对链路失效场景，本文提出了一种基于保护的可生存虚拟数据中心映射算法(Protection Based Survivable Virtual Datacenter Mapping Algorithm: P-SVDMA)，该算法通过预设链路分离的冗余链路实现对链路失效的容忍。为了尽量减少冗余资源开销，本文还提出了基于迁移保护的可生存虚拟数据中心映射算法(MigrationProtection Based Survivable Virtual Datacenter Mapping Algorithm: MP-SVDMA)。MP-SVDAMA算法的基本思想是当出现链路失效后对虚拟数据中心资源进行重新配置，将与失效链路相关的虚拟机由原来的位置迁移到一个新的可用位置，以达到绕开失效链路的目的。针对虚拟机失效的问题，本文提出了单节点冗余算法和K节点冗余算法。通过对原始的虚拟数据中心进行冗余增强，并将其映射到合理的物理资源上，确保当虚拟机发生失效时（虚拟机失效或者放置虚拟机的物理服务器故障），对外提供给租户的虚拟数据中心仍然可用，从而提高虚拟数据中心的可靠性。并且通过合理的资源配置和共享机制，提高了数据中心的资源效用。针对高可靠性虚拟数据中心的搭建问题，本文采用开源集群管理软件Pacemaker进行了基于云的高可靠性虚拟数据中心搭建。文中首先介绍了平台搭建中用到的相关技术，接着对Pacemaker的功能和特征做了简单描述，然后详细介绍了平台的搭建流程，最后分别针对虚拟数据中心应用级（MySQL）和虚拟机级的高可靠性进行实例说明。
[Abstract]:Virtual data center (VDC) for cloud computing is subject to a wide range of concerns with its low price and a very attractive, on-demand mode of payment. A growing number of businesses move from the traditional operating model of deploying dedicated servers to the resources of the leased cloud data center to meet the business's day-to-day IT application needs, thereby freeing up the high cost of maintaining and managing infrastructure networks. By virtualization technology, these tenants can efficiently share the resources of the same infrastructure and gain a more economical infrastructure as a service (IaaS). Compared with the traditional data center model, the resource virtualization technology in the cloud computing environment has a great advantage in improving the utilization ratio of resources and reducing energy consumption and the like. But its highly integrated resource organization and management approach also lead to a potential crisis. There are hundreds of data centers each year with service interruption events leading to heavy losses. At the same time, in order to reduce the cost, the cloud data center generally adopts commercial and non-telecom-level bottom-layer hardware, and the reliability of the cloud data center is a problem, and the high reliability of the virtual data center has become a problem to be solved urgently in the academic and industrial circles. The following three problems are studied in this paper: (1) The redundant design of the virtual data center when the link fails, the purpose is to ensure the availability of the virtual data center under any single-link failure scenario Sexual; (2) The redundant design of the virtual data center when the virtual machine fails (the virtual machine itself fails or the physical server of the virtual machine is hosting the virtual machine) for the purpose of achieving the availability of the virtual data center when the single virtual machine fails (3) Complete the construction of high-reliability virtual data center platform by using the open source software, Pacemaster, and realize the high reliability of application level and virtual machine level through different redundancy settings In view of the link failure scenario, a protection-based Virtual Datacenter Mapping Algorithm (P-SVDMA) is proposed in this paper. In order to minimize the overhead of redundant resources, a migration-protected Virtual Datacenter Mapping Algorithm (MP-SVD) is also proposed in this paper. the basic idea of the MP-SVDAMA algorithm is to reconfigure the virtual data center resource after the link failure, to migrate the virtual machine associated with the failed link from the original location to a new available location to achieve the bypass of the failed link To solve the problem of virtual machine failure, a single-node redundancy algorithm and a K-section are proposed in this paper. Point-to-point redundancy algorithm. By providing redundancy to the original virtual data center and mapping it to a reasonable physical resource, make sure that when the virtual machine fails (the virtual machine fails or the physical server on which the virtual machine is placed), the virtual data provided to the tenant is externally provided The heart is still available, increasing the virtual data The reliability of the heart is improved, and the data center is improved through reasonable resource allocation and sharing mechanism In order to solve the problem of high reliability virtual data center, this paper uses the open source cluster management software, Packeaker, to carry on the cloud-based high-reliability virtual data center. The paper first introduces the related technologies used in the platform construction, then briefly describes the functions and features of the Packeker, and then introduces the building process of the platform in detail, and finally the high reliability of the virtual data center application level (MySQL) and the virtual machine level respectively.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP308

【相似文献】