There are many effective heuristic algorithms for scheduling in the computational grid, however most scheduling strategies have not consider how to effectively share abundant applications software, especially commercial applications software with limited licenses. In this paper, we propose an application-oriented on-demand scheduling approach. It is based on the static and near-real-time applicationspecific and system-specific information. The scheduling system on-demand intelligently reacts to diverse applications and situations on behalf of the end-users. The implementation and the experiments results show that it delivers good performance.