A flexible

dynamic

self-adapting and scalable model is presented for managing large-scale distributed networks. A hierarchical architecture

supporting distributed management based on domain

is adopted. It achieves scalability by modifying the number of components according to the size of network. To monitoring network events efficiently

two event types are defined

primitive event and composite event. Event hierarchical processing mechanism can efficiently distribute the loads and improve the reliability of the system. Event composition approach named dynamic time windows can deal with out-of-order sequences of event arrivals due to communication delays in an efficient manner. Moreover

the system can trigger or disable management tasks automatically according to some network events.