As the existing mapping algorithms of large-scale virtual networks are of low execution efficiency and are prone to network partition when mapping virtual nodes and links ， a method of graph adjacency segmentation is proposed ， which decomposes a virtual network into several adjacent star configurations and reduces the size of large-scale virtual network evidently. At the same time ， a resource allocation model matching nodes and their adjacency links is proposed ， which enables nodes mapping to adapt links ’ resource state and makes links mapping match with the resource size of related nodes in a coordinated way. Thus ， the inharmonious operation of mapping nodes and links ， as well as the mismatching of allocating network resources ， is solved effectively.Simulated results show that the proposed algorithm reduces the mapping path length of virtual links and improves the mapping efficiency of virtual networks as well as the load-balancing performance ， and thus high acceptance ratio of virtual network requests can be achieved.