The Research On The Feature Recognition And Service Influence Of Faults In Large-Scale VoIP Network

Because of the increasing requirement of network performance and value-added services, the stability and availability of network are also increasingly important. Meanwhile, with the rapid expansion of capacity and more complexity of structure, networks usually interconnect various types of sub-networks. The network condition within faults has become more complex. Inferring the features of network faults according to network theory is difficult to meet the needs of fault management.VoIP service that requires more stringent network conditions is selected in this dissertation. With common large-scale network topologies designed, various typical models of equipment faults established, simulation of different network and faults conditions fetches the performance and flow data of end-to-end application and equipments. The contrast with corresponding data of non-fault cases sums up the general fault features on the application and network layers under different network conditions, and analyzes application-oriented law of fault location.Network performance is regarded as the temporal sequences which change over time and interrelate between states at different times. Therefore, the methods for temporal data such as statistical clustering and mining technologies can deal with network performance data processing. Thus the key performance and their characteristics summed up under fault conditions will describe the faults well.The simulation for rules related with faults figures will overcome the difficulty to obtain the sample data. The network fault features recognized from application performance such as delay, traffic and other network information may lead to more targeted and rapid network failure recovery. The combination of fault management and performance management is a commitment to reduce the burden of network management platform rather than a complete fault management system.