| With the development of computer science and communications technology, today's network rendering large, complex features, it's important to take effective management to network. Network fault management takes important role in the network management. Fault diagnosis is most important in fault management. Therefore better fault diagnosis method is particularly important for network management. After fault happened in network, because of so many alarms that can't as soon as possible to find the fault source , the paper take different diagnosis in for different stages and proposed use the network topology and event correlation based on cased-based reasoning for fault diagnosis techniques.First , use Simple Network Management Protocol to get the network topology, according to the relationship of the managed objects, to establish the association table between devices ,read the managed object's over time information in the MIB, according to the the associated table to filtering alarm with selective method .Eliminate the redundant alarms,false alarm information.Secondly,take the alarms which has been filter in to the cased-based reasoning system as input , in accordance with the largest proportion of the alarm to find same type case in the case library, find the alarm which similar case with the current alarm, calculated the distance between the current similar cases according to the information of the generate alarm object ,take the case which has smallest distance with the current alarm as the method to remove the fault ,if success ,take it in the case library for next search.Finally, through the simulation verification, this method of use the network topology and event correlation based on cased-based reasoning for fault diagnosis techniques is feasible. This method can shorten the filter alarm time to the original method's 44%, reduce waste the time of analysis redundant alarms , contest valuable time to remove the fault, the experiment results show our approaches are promising. |
PDF Full Text Request