面向未来网络的网络虚拟化资源管理架构与映射方法研究

发布时间：2018-02-08 09:15

本文关键词： 未来网络网络虚拟化虚拟网络映射资源管理人工鱼群 Katz中心度　出处：《哈尔滨工程大学》2016年博士论文　论文类型：学位论文

【摘要】：随着计算理论和网络技术的蓬勃发展,新型网络形态不断涌现,网络在功能形态上已经开始由“网络中心体系”向“应用中心体系”转变,一切皆服务XaaS逐渐成为未来网络发展的终极目标。当前互联网架构缺乏对复杂环境和新型应用的适应能力和应变能力,无力满足服务的多样性需求。作为革命式未来网络的核心,网络虚拟化技术能够在共享公共物理网络基础之上支持多个异构虚拟网络,并能够根据动态变化的用户需求对整个网络中的节点资源和链路资源进行合理配置,带来服务模式变革的同时满足用户服务的多样化需求。由于使用网络虚拟化技术构建未来网络体系结构提出时间较短,相关研究尚处于初级阶段,如何对分布、异构、自治、动态的底层网络资源进行抽象形成可供统一调配和管理的虚拟资源,并通过合理的映射方法构建虚拟网络以满足用户的服务需求,是网络虚拟化亟待解决的问题。据此,本文以网络虚拟化环境下资源管理架构与映射方法为研究目标,以认知科学、群体智能、拓扑学、社会网络中心度等相关理论为手段,建立一个层次化认知资源管理架构,从高效、可靠和节能三个方面提出虚拟网络映射新方法,为网络虚拟化技术的完善和发展提供参考。本文的主要研究内容如下:首先,在网络虚拟化环境下资源管理架构的构建方面,提出一种具有“资源管理集中式,任务调度分布式”特点的层次化认知资源管理架构HCRMF。层间划分为不同的管理域,域内采用集中式管理,使得HCRMF架构具有良好的可扩展性、灵活性、自治性和可靠性,降低管理难度和开销。而任务调度服务器内嵌认知单元并通过协作进行任务分配,能够减少人的干预程度,高效地应对用户的虚拟网络请求,满足用户的个性化需求。与现有集中式和分布式管理架构相比,HCRMF架构能够显著降低管理节点和链路的压力,提升虚拟网络请求的响应速度。其次,为了解决现有虚拟网络映射算法效率低和对映射问题空间有限制等问题,提出一种基于人工鱼群的启发式高效虚拟网络映射算法VNE-AFS。在底层网络资源有限和不支持路径分割的前提下,将虚拟网络映射问题描述为二进制组合优化模型,利用人工鱼群算法较强的寻优能力对虚拟网络进行近似最优分配。实验结果表明,随着时间的增长和虚拟网络请求的增多,该算法在映射成功率、平均收益上较传统的映射算法有较为明显的提升,并且有效地降低了底层网络的平均花费和求解时间。再次,为了解决因底层节点故障导致虚拟网络运行环境不可靠问题,提出一种虚拟网络可生存的启发式可靠映射算法RHM-SVN。借鉴WDM光网络路由可生存思想,确保在底层网络单节点失效情况下,被映射的虚拟网络剩余部分仍然保持连通,最大程度保证虚拟网络的完整性和服务的连续性。以最小化底层网络映射开销为目标函数,建立虚拟网络可靠映射的整数线性规划模型,并使用RHM-SVN算法进行求解。实验结果表明,该算法能够有效提高映射成功率、底层网络平均收益和虚拟网络恢复成功率。最后,为了解决当前网络依据峰值设计导致电力资源利用率较低问题,提出一种基于Katz中心度的启发式节能虚拟网络映射算法Katz-VNE。选用Katz中心度表示节点的影响力,以最小化底层网络能耗为目标,建立节能虚拟网络映射的整数线性规划模型,并使用Katz-VNE算法进行求解。该算法能够尽量将虚拟网络节点整合映射到底层网络Katz中心度高的已经工作节点上,并通过最短路径提供最佳带宽资源因子避免底层网络出现过热节点。实验结果表明,该算法能够显著地降低底层网络长期平均能耗,提高了底层网络长期收益开销比、虚拟网络请求接受率、休眠节点和链路比例。
[Abstract]:With the rapid development of computing theory and network technology, the emerging network configuration, the network has begun to change from "network center system" to "center application system" in the function form, all XaaS service has gradually become the ultimate goal of network development in the future. The current Internet architecture of the complex environment and lack of adaptability and application model strain capacity, unable to meet the needs of diverse services. As the core of revolutionary future network, network virtualization technology to share on the basis of public physical network to support multiple heterogeneous virtual networks, and to the rational allocation of the resource nodes and link resources in the network according to the dynamic changes of the needs of users, service mode meanwhile, to meet the diverse needs of service users. Due to the use of network virtualization technology to build the future network architecture is proposed For a short time, the related research is still in the initial stage, how to distributed, heterogeneous, autonomous, bottom cyber source dynamic abstract form the virtual resources available for unified deployment and management, and through the reasonable construction of virtual network mapping method to meet the requirements of customers, is the urgent need to solve the problem of network virtualization. Accordingly. In this paper, the network virtualization environment resource management architecture and mapping method as the research object, in cognitive science, swarm intelligence, network topology, the center for social degree and other related theory as means, establish a hierarchical knowledge resource management architecture, from the high, the paper puts forward a new method of virtual network mapping three aspects to provide reliable and energy-saving. The reference for the improvement and development of network virtualization technology. The main research contents of this paper are as follows: firstly, in the construction of infrastructure resource management in the network virtualization environment, put forward a "Centralized resource management, task scheduling of distributed characteristics of hierarchical cognitive resource management architecture HCRMF. layer is divided into different administrative domains, using a centralized management domain, the HCRMF framework has good scalability, flexibility, autonomy and reliability, reducing management difficulty and cost. And the task scheduling server embedded cognition unit and through collaborative task allocation, can reduce the degree of human intervention, the virtual network efficiently respond to user requests, meet the personalized needs of users. Compared with the existing centralized and distributed management architecture, HCRMF architecture can significantly reduce the management node and link stress, improve the response speed of virtual network requests. Secondly, in order to solve the mapping the existing algorithms of virtual network mapping problem of low efficiency and space restrictions and other issues, put forward a kind of efficient heuristic based on virtual artificial fish Virtual network mapping algorithm VNE-AFS. in the underlying cyber source limited and does not support the premise of path segmentation, will describe the virtual network mapping problem for binary combinatorial optimization model of the virtual network to approximate optimal allocation based on artificial fish swarm algorithm strong searching ability. The experimental results show that with the increase of the growth time and virtual network requests. The success rate of the algorithm in the mapping, compared with the traditional mapping algorithm average returns have obvious improvement, and effectively reduces the average cost of the underlying network and the solution time. Again, in order to solve the bottom node failures lead to the virtual network environment is not reliable, RHM-SVN. proposed a reliable heuristic mapping algorithm of virtual network survivability from the routing WDM optical network survivability, to ensure that in the case of single node failure of the underlying network, virtual network mapping of the remaining part Points remain connected to ensure the maximum degree of integrity and continuity of virtual network service. In order to minimize the underlying network mapping cost as the objective function, an integer linear programming model to establish a reliable virtual network mapping, which is solved by RHM-SVN algorithm. The experimental results show that the algorithm can effectively improve the success rate of the underlying network mapping, average income the virtual network and the rate of recovery. Finally, in order to solve the current network based on peak design leads to lower utilization of power resources, this paper puts forward a Katz center of such virtual network mapping heuristic algorithm based on Katz-VNE. with Katz center of the said node influence, to minimize the network energy consumption and integer linear programming model for the establishment of energy-saving the virtual network mapping, which is solved by Katz-VNE algorithm. The algorithm can try to reflect the integration of virtual network node Has been working for the underlying network node shoot Katz center high degree, and provide the best bandwidth factor to avoid overheating of the underlying network node through the shortest path. The experimental results show that this algorithm can significantly reduce the long-term average energy consumption of the underlying network, improve the long-term gains in cost than the underlying network, virtual network request acceptance rate, node dormancy and the proportion of the link.

【学位授予单位】：哈尔滨工程大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.0

【相似文献】