Abstract: this paper proposes a cross-domain virtual network embedding algorithm based on dynamic topology perception and resource attributes in the multi-domain converged network architecture. By analyzing the topological property of nodes within virtual network and physical network from the local and global perspective, we establish a node multi-attribute evaluation model combined with the expand network resource. Based on this model, the principal component analysis method and TOPSIS (technique for order preference by similarity to an ideal solution) are introduced to measure the embedding priority of nodes. Then the network load condition is analyzed with link resource price. The simulation results depicts that this algorithm improves the construction success rate of multi-domain virtual network requests, increases the network revenue to cost ratio, and could reduce the network embedding delay.