Research Of Virtual Network Embedding Technology For Network Virtualization

The Internet has become one of the most important infrastructure of human society, which is profoundly influencing and changing our life. However, with the rising of network scale and the rapid development of network business applications, the traditional Internet is facing more and more challenges in data security, network management, quality of service and so on. To overcome these problem, researchers attempt to promote the innovation of network technology and architecture by building the experiment platform in network virtualization environment. And the core idea of network virtualization is creating and running several different virtual networks on the sharing physical network at the same time by using virtualization technology.Virtual network embedding problem is one of the core issues in the study of network virtualization technology, which aims to successfully and efficiently assign the virtual network requests onto a shared substrate network. On the basic of systematic learning of the virtual network embedding problem, this paper is mainly aiming at researching of the shortcomings in existing solutions, including the mapping efficiency, success rate of virtual network mapping and the optimization process for different requirement of virtual network. To form a efficient, flexible and scalable mapping solution, this paper has designed three different virtual network embedding algorithms and summarized some common characteristics during the embedding. In this paper, the main innovation work includes the following three aspects:(1) Aiming at improving the efficiency of the virtual network mapping, this paper proposes a new virtual network embedding algorithm based on the node connectivity. The existing word mainly consider the nodes’local resource, such as node calculation ability and export bandwidth, or the resource of neighbor nodes during the design of heuristic embedding algorithm, which ignore the topological characteristics between nodes in network. To address this issue, this paper proposes a connectivity evaluation model for by considering the information of both the distance and available bandwidth between each pair of nodes in network, and then puts forward a two stage algorithm for virtual network embedding. The simulation results show that the virtual network mapping algorithm based on node connectivity achieve higher virtual network acceptance ratio and better revenue/cost ratio, while reducing the running time of the algorithm.(2) Aiming at improving the success rate of the virtual network mapping, this paper proposes a new virtual network embedding algorithm based on the real-time topological attributes. Most of the exiting work use the same methods to measure the virtual nodes and physical nodes, without considering the different topological characteristics between virtual network requests and underlying physical network in the mapping system which may pose significant impact on the performance of the embedding. To address this issue, this paper introduces the node centrality theory which is already widely used in study of social network and redefined the node centricity measurement model based on the background of the virtual network embedding problem to evaluate the virtual nodes and physical nodes in different ways, and then puts forward a two stage algorithm for virtual network embedding. Since this new algorithm not only reflect the local resource state of nodes, but also consider the real-time state of node centrality in the process of node mapping, virtual nodes are allocated to the more appropriate physical nodes and more virtual network requests are mapped successfully. The simulation results show that, under the different condition of virtual network resource requirements, the virtual network mapping algorithm based on real-time topological properties achieve higher success rate of virtual network mapping, and get a good revenue/cost ratio and algorithm running time(3) Aiming at improving the performance of embedding with different virtual network requirements, this paper proposes a new complex algorithm with topology recognition for virtual network requests in multiple topologies. In the business system of network virtualization, service provider build virtual network requests in different topology structures based on the needs of network application to provide customized Internet services. In order to get closer to the practical business system and achieve better mapping scheme according to the different topological characteristics of virtual network requests, this paper presents the design idea of the embedding algorithm for virtual network requests in different topologies, and then puts forward a general evaluation model for physical nodes by considering nodes’local resource and the topological connections in the network to help measuring the importance of physical nodes. Based on the above work, this paper design a complex algorithm with topology recognition for virtual network requests in multiple topologies. The simulation results show that the proposed complex virtual network mapping algorithm improve the requests acceptance rate and revenue/cost ratio in multiple topologies environment, which significantly improve the performance of the virtual network embedding.