云服务的部署及管理机制研究

发布时间：2018-06-20 12:27

本文选题：云计算 + 云服务　；参考：《北京邮电大学》2014年博士论文

【摘要】：云计算作为一种新型的计算模式能够以服务的形式按需向用户提供弹性可扩展的资源,从而大幅度提高资源利用率、降低信息服务成本。而云服务可靠提供的关键在于云服务的部署及维护管理,即如何为云服务合理分配资源以保证云服务的高效部署及可靠运行。然而,云服务部署中存在的用户多样性、基础设施的海量异构性等特点,使得云计算平台(简称云平台)在云服务部署方面遇到了诸多挑战,例如,如何进行虚拟镜像的有效获取、如何合理的分配资源,以及如何在服务器在线维护情况下执行资源的动态调整等。因此,针对上述问题,对云服务的部署与管理机制进行了深入研究,具体研究内容如下： (1)基于多点协作的虚拟镜像文件高效获取方法。针对在IaaS云按需型资源提供方式下,大尺寸虚拟机模板文件获取耗时长导致的用户虚拟机请求部署效率低下问题,提出了一种基于多点协作的虚拟镜像文件动态获取方法。该方法首先根据数据中心镜像文件的存储位置及镜像文件本身的特点,设计一种混合分块策略,用于将虚拟机模板文件划分成细粒度的块文件,然后,基于细粒度块的本地缓存,提出了一种多点协作的虚拟机模板文件高效获取方法,最后,仿真实验表明,该方法能够快速响应用户所需的虚拟机模板文件请求,实现虚拟机的高效部署。 (2)基于应用流友好的虚拟镜像文件传输方法。针对在IaaS云预约型资源提供方式下,大尺寸虚拟镜像文件传输给数据中心网络造成大的额外网络流量负载导致的数据中心应用性能急剧下降问题,提出了一种应用流友好的虚拟镜像文件传输方法。该方法首先将随时间变化的数据中心网络建立在时间扩展图上,以虚拟模板文件传输给数据中心网络带来的负载影响最小为目标,建立时延容忍的虚拟机模板文件传输的问题求解模型,然后,针对该问题模型,采用遗传优化算法进行最优求解,最后,仿真实验表明,该方法能够充分利用数据中心网络上空闲的或是负载较轻的链路,为数据中心网络上承载的各种应用提供一个更为均衡的网络,从而减少大尺寸虚拟镜像文件传输对数据中心网络上承载的各种应用的性能影响。 (3)面向负载均衡的虚拟网络资源分配方法。针对虚拟网络资源分配中物理路径上中间节点资源瓶颈而导致后续虚拟网络请求成功率降低问题,提出一种负载均衡的虚拟网络资源分配方法。该方法首先考虑物理路径上中间节点资源消耗,以节点负载和链路负载同时达到均衡为目标,将路径跳数限制作为约束,建立虚拟网络资源分配的问题求解模型,然后,针对该问题模型,采用多目标负载均衡粒子群优化算法求解,最后,仿真实验表明,该分配方法能够有效消除资源瓶颈(为后续虚拟网络请求提供了一个更为均衡的底层物理网络)从而提高虚拟网络构建成功率、网络资源利用率以及基础设施提供商的收益。 (4)服务器在线维护下通信成本感知的资源调整方法。针对数据中心服务器在线维护时虚拟机迁移而导致虚拟网络通信开销增加问题,提出了一种服务器在线维护下通信成本感知的资源调整方法。该方法首先找出服务器在线维护情况下虚拟机迁移策略的候选集,并以虚拟网络中虚拟机之间的通信开销最小为目标,以物理服务器的资源能力为约束,建立虚拟机迁移的问题求解模型,然后,针对该问题模型,采用模拟退火算法求解,最后,仿真实验表明,该调整方法能够有效避免虚拟机由于服务器维护而造成的多次迁移,减少虚拟机迁移给网络带来的迁移流量开销,减少多层应用中虚拟机之间的通信开销。
[Abstract]:As a new computing mode , cloud computing can provide users with flexible and scalable resources in the form of service , thereby greatly improving resource utilization rate and reducing information service cost . The key to reliable provision of cloud services lies in the deployment and maintenance management of cloud services , namely , how to allocate resources reasonably for cloud services to ensure efficient deployment and reliable operation of cloud services .

However , the cloud computing platform ( simply called the cloud platform ) meets many challenges in cloud service deployment , such as how to get effective acquisition of virtual mirror image , how to allocate resources reasonably and how to perform dynamic adjustment of resources under the condition of online maintenance of the server . Therefore , the deployment and management mechanism of cloud service is researched deeply , and the specific research contents are as follows :

The method comprises the following steps : firstly , according to the storage position of a data center image file and the characteristics of the image file , a mixed blocking strategy is designed to divide the virtual machine template file into a fine - grained block file , and finally , based on the local cache of the fine granularity block , a multi - point cooperative virtual machine template file high - efficiency acquisition method is proposed , and finally , the simulation experiment shows that the method can quickly respond to the request of the virtual machine template file required by the user and realize the efficient deployment of the virtual machine .

The invention relates to a virtual image file transmission method based on application flow friendliness , aiming at solving the problem that the data center application performance caused by large extra network traffic load caused by large - sized virtual image files to a data center network in the mode of providing a large - sized virtual image file to a data center network is extremely reduced .

The method comprises the following steps : firstly , taking into account the resource consumption of an intermediate node on the physical path , taking the node load and the link load as a target , solving a problem solving model of the virtual network resource allocation , and finally , aiming at the problem model , adopting a multi - target load balancing particle swarm optimization algorithm to solve the problem model , and finally , simulating experiment shows that the distribution method can effectively eliminate the resource bottleneck ( providing a more balanced bottom layer physical network for subsequent virtual network requests ) , thereby improving the success rate of the virtual network construction , the utilization rate of network resources and the benefits of the infrastructure providers .

The method comprises the following steps : firstly , finding a candidate set of a virtual machine migration policy under the condition of online maintenance of a data center server , establishing a candidate set of the virtual machine migration strategy under the condition of online maintenance of the server , establishing a problem solving model of the virtual machine migration by using the resource capacity of the physical server as a target , and finally , simulating an experiment to show that the adjusting method can effectively avoid the multiple migration caused by the maintenance of the virtual machine , reduce the migration flow cost caused by the migration of the virtual machine to the network , and reduce the communication cost between the virtual machines in the multi - layer application .
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】