In today’s cloud environment, the on-demand leasing of virtual resources can provide great flexibility for data centers, especially the resource leasing with the granularity of virtual network can provide much better personalized demand support for users. To allocate reasonable physical host and network resources for user’s requirement including nodes and links is called Virtual Network Embedding (VNE) problem. Most of the existing VNE algorithms are general algorithms on random topology, and are not optimized for the topological structure of data centers. Therefore, the efficiency and optimization extent need be improved in dealing the VNE problem in data centers. According to the characteristics of data center topology, a virtual network embedding algorithm based on node connectivity ranking named BS-VNE is proposed. Firstly, a maximum spanning tree algorithm is designed to sort the virtual nodes. The sorting algorithm calculates the node connectivity according to both the bandwidth and connectivity of virtual nodes and the connectivity of virtual nodes in the whole network, so as to obtain more reasonable ranking results. Then, a discrete particle swarm optimization algorithm is used to solve the mapping solution of virtual networks according to the results of virtual node ranking result. In the process of solution searching, the heuristic rules for physical network topology of data center are introduced and combined into particles search process to improve the solution efficiency. Simulation results show that the proposed algorithm can improve the benefit/cost ratio and resource utilization ratio of physical network.