Event-Driven Based Hierarchical And Adaptive Network Management System

Nowadays network management solutions is facing the problem of lacking aunified management processes, including functional and business processes. Anotherproblem is that SNMP-based network management systems usually use a centralizednetwork management model which is not suitable for large-scale network system. Tosolve the above problems,a distributed event-driven network management systemXNMS is constructed and implemented.The system is based on event-driven model which has strong scalability infunction,enabling massive network events to be buffered, filtered and distributed; Tosolve the performance bottleneck of centralized architecture, three layers adaptivedistribution system making use of ICE technology is designed and implemented. themanagement structure is visualized by Applet technology to fullfill remote adjustmentof the management structure. A topology-sensed task allocation solution is alsoproposed. First of all, the Floyd-Warshall algorithm is used to calculate distancebetween nodes. Greedy-Walking is proposed to find the location of central managementnode.Then the minimum distance hierarchical clustering algorithm is used to achievetopology-aware clustering; when the system is running, the load of management node ismonitored to adjust for clustering in time. Experiments show that, XNMS can unifynetwork management process, solve the performance bottlenecks of centralizedmanagement, and has strong feasibility in task allocation.