Research On Survivable Virtual Network Embedding Algorithm In Software Defined Network

With the rapid development of cloud computing and the expasion of data center network multi-tenant requirements, higher requirements for the management and distribution of network resources is put forward. Virtual network embedding algorithms can realize the resource mapping of the virtual network on the substrate network, which effectively implements the management and distribution of substrate network resources. Among the virtual network embedding algorithms, the default resources allocated to the tenant can always work effective without any failure.Yet in reality,equipment performance problem may cause the failure of substrate equipment or connection failure. In order to guarantee the reliability of the tenants' virtual network services, it is necessary to put forward some survivable virtual network embedding algorithms based on the virtual network embedding algorithm.Most of the existing survivable virtual network embedding algorithms are based on traditional data center network, with less research on the emerging Software Defined Network (SDN). SDN separates control and forwarding function, with the characteristics of centralized control,network virtualization, open interface and so on, which can well meet the demand of network resource centralized management in multi-tenant data center network and provide better management and control methods during the resource allocation process of virtual network embedding as well as recovery, so it is very valuable to do the research based on SDN.The existing survivable virtual network embedding (SVNE)algorithms mainly provide two kinds of failure recovery strategy, active protection and passive recovery strategy. However, both of strategies are not good enough. Active protection strategy backups the virtual network before embedding, which chooses the important nodes of virtual network as the backup nodes. In the process of embedding, not only the mapping resources of nodes but also the backup resources of all the backup nodes need to be provided, otherwise the vitural network request (VNR) will be rejected, which will lead to a low request acceptance rate. Passive recovery strategy needs to use the existing resources to restore the failure when the failure occurs, which will lead to a low failure recovery efficiency and great fluctuation of failure recovery success rate (depends on the remaining substrate network resources).In order to improve the survivability of vitural network, we proposed a survivable virtual network embedding algorithm in the SDN network environment. The algorithm adopts resource-aware backup strategy based on the existing failure recovery strategies. The algorithm tries to provide all the required resources of virtual network. The algorithm will detect the remainning substrate network resources before supplying the backup resources to the virtual network. While it only supply the backup resources to those virtual nodes which match the available resources when the substrate network can not provide all the resources. Those virtual nodes without backup resources can use the remaining substrate network resources to restore the failure when node failure occurs. Compared with the active protection strategy which rejects the virtual network request without the required backup resources, the proposed strategy can improve the request acceptance rate. Compared with the passive recovery strategy,it in this paper can use not only backup resources but also the remaining substrate network resources to restore the failure, which can improve the failure recovery efficiency and success rate. At the same time, the concept of location-constrained is introduced in our strategy. In the process of virtual node mapping and backup mapping, we need to ensure that both the physical nodes for virtual node mapping and the backup mapping nodes meet the location constraints. In this way, after the failure recovery, the nodes in the virtual network will not have long connection and loose structure problem.In the paper, we choose two kinds of survivable virtual network embedding algorithms for comparision, namely the active protection strategy and passive recovery strategy. SDN is selected as the substrate network. We compare and analyze the performance of the algorithm from eight aspects, namely virtual network request acceptance rate, unit time revenue, average cost of per VNR, revenue/cost ratio, substrate resource utilization ratio, failure recovery ratio and algorithm time. The simulation experimental results show that the algorithm proposed in this paper effectively improves the virtual network request acceptance rate, failure recovery efficiency and success rate on the basis of a good revenue, high substrate resource utilization ratio and acceptable cost.Firstly, this paper analyzes the current research situation and the problems of existing survivable virtual network embedding; then it introduces the proposed survivable virtual network embedding algorithm in software defined network in detail; finally, this paper introduces the design and realization of simulation platform, and the performance of the proposed algorithm is evaluated.