Research On Cost-revenue Ratio Optimization For Large-scale Virtual Network Mapping

With the rapid development of cloud computing,virtualization technology,as its core technology,is becoming more and more important.As one of the most promising branches of virtualization,network virtualization has also developed rapidly recently.The virtual network mapping divides the network request into many independent virtual networks,and then the virtual network is mapped to the physical network shared by the all virtual networks to realize the customized service of various kinds of users.On such an underlying physical network,a number of virtual networks can be run simultaneously to achieve an end-to-end service.There has been a lot of papers to study the problem of virtual network mapping,or virtual network embedding.The main research strategy is divided into two kinds: One-Stage strategy and Two-Stage strategy,the main difference is whether to Merge virtual node mapping and virtual link mapping,which are two independent sub processes.However,at the present stage,when the size of the virtual network requests becomes larger,many One-Stage algorithms have the problems of large time complexity.And because of the connection between virtual nodes and links,considering the constraints of virtual nodes and links at the same time,the processing is relatively complicated.Similarly,many Two-Stage algorithms are based on smallscale virtual network mapping.When the size of the virtual network becomes larger,the success rate of the algorithm is low,which leads to the low revenue of the algorithm,and the bandwidth cost of the virtual link mapping is also very large.The contribution of this paper is mainly in two aspects.On the one hand,according to the topology of real network and the development trend of network virtualization.we put forward a new definition of large-scale virtual network mapping.The main improvement point is to expand the size of the virtual network,and to improve the mapping limit of each virtual node.A new problem definition based on optimized cost gain ratio is proposed.On the other hand,considering the main challenge is how to reasonablely map virtual nodes in the virtual network mapping problem,we put forward two kinds of new algorithm:a Link-Opt algorithm,a Node-Merge algorithm under the definition of new problem.The former is a One-stage algorithm,the algorithm is mainly to optimize the bandwidth mapping cost,and the process of virtual node mapping is distributed in the process of link mapping.The latter is a Two-Stage algorithm.In the process of node mapping,the method of node merging simplifies the virtual network topology,and optimizes the traditional shortest path routing algorithm in the process of link mapping.Finally,we set up a virtual network mapping system carried out several simulation experiments,experiments show that the optimization algorithm we propose does optimization Cost-Revenue ratio,improving the mapping acceptance rate.