Present web service selection method is focusing on single user request and selects service with best QoS attributes. However, conditions that multiple user requests for the same functional service are lack of consideration. In such situation, service with best QoS will be selected by many users, causing conflict and overload. This paper aims at selecting services from a global view for all users, thereby optimizing service resources utilization and improving users' satisfaction. It first normalizes different QoS attributes into one single value. Then a global optimal service selection model is raised and calculated by bipartite graph method with some modifications. Simulation results demonstrate that the method can find an optimal selection solution in a very fast way.