The rapid development of Internet makes network management on large-scale network a critical issue. But with the management task of large-scale network becoming more complicated, neither centralized network management nor agent based network management can satisfy the increasing demands. This paper presents a network management framework to support dynamic scheduling decisions. In thisframework, some algorithms are proposed to decompose the whole network management task into several groups of sub-tasks. During the course of decomposition, different priorities are assigned to sub-tasks. Then based on the priorities of these sub-tasks, the strategies of agent scheduling are established. Priority-ranked sub-tasks are grouped according to their inter-dependences. Sub-tasks with the same priority are put info fhe same group and they can be performed in parallel manner, while different groups of sub-tasks with different priorities must be implemented according to the order of their priorities. An experiment has been done with the algorithms, the results of which demonstrate the advantage of the algorithms.