Research And Realization Of The Network Fault Management System Based On Dependence Relationship

With Computer network providing a wealth of business, people are becoming increasingly dependent on it and raising their requirements on the network stability and availability. The task of network fault management is to detect network failure on time and provide treatment recommendations to reduce the difficulty of troubleshooting through monitoring the network state. This article focuses on the techniques of building the state data collection and processing and fault merging, aimed to provide a reasonable and efficient management system scheme.First, the paper introduces some standard protocols for network management, based on which most equipment manufacturers provide the running parameters and access interface. Those protocols are the main source of status data. While, some common alarm correlation techniques are illuminated.From the perspective of fault management, a fault is caused by the state change of a physical device and manifested by a certain degree of symptoms. Therefore, a reasonable data collection and processing on the attributes parameters of network resources is very important to the timely detection of failures. There are two main methods for data collection: Trap informs and active polling. Some techniques are presented to improve the collection efficiency. In order to use the wealth of enterprise customized trap information, this paper focuses on setting the rules RMON through Telnet and SSH communication; in order to solve the trap transmission problems caused by information loss, it take use of proactive detection methods to maintain state consistency. When the event arrives, based on the event distribution strategy of events-interest associated, the case will be passed to the interested managed object, to carry out its own state or polling strategy adjustment.Then, based on the natural characteristics of computer networks, the paper builds a fault propagation model based on the layered dependence relationship beteewn nework entities. The model starts from the view of the data flow transmission between entities in the network to avoid the complexity of modeling, and it can match SNMP MIB design idea, you can take full advantage of the MIB variables reflecting the equipment operating status information, considering the fault correlation caused by the network topology, and able to adapt to topology changes.According to the diffirent of alarms cause, two ways of faults merging are suggested. The emerging solution of hardware alarm is based on equipment, and performance alarm base on service. This way can help an administrator extract the most likely faults from a large number of alarms, which should be given priority to diagnosis.Finally, this article designs the detailed modules of fault management system and accomplishs the programming. The system provides a friendly user interface using Visio-based graphical control, reminds the fault through the color change and the alarm voices, and gives faults report for the administrator to view the faults history.