Nowadays, Internet plays an irreplaceable role in almost every aspect of human life. However, it is also the popularity and increasing limitations which impede its development. Future Internet requires more flexible and more scalable architecture to support multiple parallel applications. Thus, how to select a suitable architecture at run-time to obtain the system optimum is a key problem in future Internet. Network Virtualization is supposed to be a key technology of future network, which uses different network protocols and supports different network applications in a common physical substrate. This paper presents a parallel selection algorithm to dynamically select the suitable network architectures for multiple applications coming from user level simultaneously in network virtualization. First, we determine the weight of each application based on analytic hierarchy process (AHP). Then we calculate the utilities of each candidate in parallel. Finally, we present our selection algorithm based on a modified Kuhn-Munkres algorithm. The simulation result shows that the selection algorithm obtains a better performance than the other two algorithms and the system optimum is achieved.