Research On Mapping Algorithm Based On Node Clustering For Data Center Network Virtualization

In recent years,with the rapid development of modern society,computer network has made a great progress,and the Internet has become an integral part of the way people work,live and learn. The Internet has become a strong support for modern social economic development and improving the quality of life,and accelerating the pace of development of science and technology. However,with the rapid increase in the number of users,the Internet’s pressure increasing. The traditional architecture of the Internet arises more and more problems,and nowadays“best-effort”Internet service model on security,stability and quality appeared to be inadequate,and there is also a severe ordeal of technology. This idea seriously affected the production and development of innovative network technologies,and does not solve the current problems and deficiencies in the Internet fundamentally. Faced with this problem,in recent years,data centers are becoming central information system for enterprises and tenants,and provides a good environment for them. Data centers in all sizes are emerging, the telecommunications industry, financial institutions, government agencies,are gradually improving their own data centers,and small businesses are constantly building their own data centers. Data center network virtualization technology has been widespread concerned in academia and industry,network virtualization technology makes it possible to implement multiple heterogeneous virtual networks on the same underlying facility. Virtual network is composed of virtual nodes and virtual links. Since the underlying physical network is programmable,virtual network can run many protocols like IP on the underlying physical network with certain topology, and therefore, network virtualization technology can theoretically achieve the deployment of new network protocols,architecture and applications under the premise of existing network. Network virtualization technology provides a feasible way to the evolution of the Future Internet and should be one of indispensable features of the Future Internet. With the gradually development and maturation of software-defined network(SDN),the implementation of network virtualization not strongly rely on physical network devices any more,so the network virtualization’s “boundary”has also been greatly improved.Rapid development of cloud computing and large demand of network operators need the strong support of the underlying data center. Therefore,the effective using the substrate resources and achieving maximum revenue are becoming the target of the operators. Our objective is to provide an innovated source allocation algorithm during the process of data center network virtualization,which is able to guarantee the stability and scalability for heterogeneous data centers,and obtain the best revenue as well. In this paper,we first formulate a virtualization model based on variational inequality. It aims at the maximum long-term average revenue,and ensures load balancing for substrate network. Based on this model,we propose a node clustering resource allocation algorithm. We compare our proposal with a recent representative resource allocation algorithm through a series of simulations. Our simulation results show that it can greatly increase the acceptance ratio,resource utilization and long-term average revenue.