面向网络虚拟化的虚拟网映射方法研究

发布时间：2018-07-03 11:47

本文选题：网络虚拟化 + 虚拟网映射　；参考：《福州大学》2014年硕士论文

【摘要】：通过网络虚拟化的思想破解当前互联网出现的僵化问题已是大势所趋。近些年,软件定义网络(Software Defined Networking, SDN)的提出,大大推动了网络虚拟化的发展。如何从物理网中划分出虚拟网即虚拟网映射问题,是网络虚拟化的核心问题。经过多年的研究,虚拟网映射算法已经有了长足的进步,但是仍存在以下问题：对于规模较大的虚拟网请求,映射效率不高；传统的虚拟网映射方法不能满足跨区域分布式并行构建的要求。针对这些问题,本课题通过对底层网络节点的局部感知和虚拟网链路映射阶段的研究,分别提出基于场强引力的局部拓扑感知虚拟网映射算法和链路均衡的双向搜索算法,并基于二者思想提出分布式并行虚拟网映射框架与算法。具体而言,本文工作包括以下三个部分：(1)基于场强引力的局部拓扑感知虚拟网映射算法,通过节点间的相互作用力计算节点的重要性排序,有利于实现虚拟网映射的选择过程。在协同节点映射和链路映射基础上,该算法将拓扑势描述为虚拟网中节点受自身和邻近节点作用力的共同影响,通过局部拓扑感知的方式实现节点排序,并建立已映射节点与待映射候选节点集间的场强引力模型。该模型能够控制节点的影响范围并反映节点受局部网络拓扑的影响,通过直接计算的方式快速获知节点排序信息,有效地克服了全局拓扑信息迭代方式更新频繁且通信开销大的问题。(2)针对虚拟网需求映射方法中复杂度较高和效率较低的问题,结合源节点和汇节点并行的带宽优先策略,提出一种链路均衡的双向搜索算法。该算法减少搜索层次,缩小了搜索空间,提高了算法的效率；同时保证底层网络链路的均衡性,从而提高虚拟网映射的成功率。(3)为充分利用路由节点的计算能力,实现节点的虚拟化与智能化,基于局部拓扑感知和双向搜索思想,本课题提出了分布式并行虚拟网映射算法,设计了一种基于MapReduce的节点自治和节点间平衡对等的智能分布式映射框架,并实现基于信息素传播和消息通信的分布式在线虚拟网映射算法。该算法框架将虚拟网请求分成不同的元组,通过底层路由节点分发元组并收集元组构建结果,从而完成虚拟网分布式并行映射。本课题的研究成果为虚拟网请求能够在底层网络中分布式并行部署提供了理论依据和技术支持,同时也为虚拟网自愈的研究提供了一定的参考价值。
[Abstract]:It is the trend of the times to solve the current rigid problem of the Internet through the idea of network virtualization. In recent years, the development of network virtualization has been greatly promoted by the development of Software defined Network (SDN). How to divide virtual network from physical network is the core problem of network virtualization. After years of research, virtual network mapping algorithm has made great progress, but there are still some problems: for large-scale virtual network requests, mapping efficiency is not high; The traditional virtual network mapping method can not meet the requirements of cross-regional distributed parallel construction. In order to solve these problems, through the research of local perception and link mapping stage of the underlying network nodes, this paper proposes a local topology sensing virtual network mapping algorithm based on field strength gravity and a two-way search algorithm for link equalization, respectively. Based on the two ideas, the mapping framework and algorithm of distributed parallel virtual network are proposed. Specifically, the work of this paper includes the following three parts: (1) based on the local topology sensing virtual network mapping algorithm, the importance ranking of nodes is calculated by the interaction between nodes. It is propitious to realize the selection process of virtual network mapping. On the basis of cooperative node mapping and link mapping, the algorithm describes the topological potential as the joint influence of the forces of the nodes themselves and adjacent nodes in the virtual network, and realizes the sorting of nodes by means of local topology perception. A model of field strength between mapped nodes and candidate node sets is established. The model can control the influence range of nodes and reflect that nodes are affected by local network topology. It effectively overcomes the problem of frequent updating of global topology information and high communication overhead. (2) aiming at the problem of high complexity and low efficiency in the requirement mapping method of virtual network, the bandwidth priority strategy of parallel between source node and sink node is combined. A bidirectional search algorithm for link equalization is proposed. The algorithm reduces the search level, reduces the search space, and improves the efficiency of the algorithm. At the same time, it ensures the equilibrium of the underlying network link, thus improves the success rate of the virtual network mapping. (3) in order to make full use of the computing power of the routing node, In order to realize the virtualization and intelligence of nodes, based on the idea of local topology perception and bidirectional search, a distributed parallel virtual network mapping algorithm is proposed in this paper. In this paper, an intelligent distributed mapping framework based on MapReduce, which is based on node autonomy and balanced equivalence between nodes, is designed, and a distributed online virtual network mapping algorithm based on pheromone propagation and message communication is implemented. The algorithm framework divides the virtual network request into different tuples, distributes the tuples through the underlying routing nodes and collects the construction results of the tuples, thus accomplishing the distributed parallel mapping of the virtual networks. The research results of this paper provide a theoretical basis and technical support for distributed parallel deployment of virtual network requests in the underlying network, and also provide a certain reference value for the research of virtual network self-healing.
【学位授予单位】：福州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01

【相似文献】