Research On Fault Diagnosis And Localization Techniques In Complex Information Networks

Network fault diagnosis is an important part of network management, which is essential to ensure the quality of service for the network. In this paper, the common network fault diagnosis methods in recent years are summarized. These methods are divided into detection technology and data processing two categories, the applicability and advantages of these methods are compared. However, the development of the network has put forward new requirements to the technology of fault diagnosis, the future work and improvement ideas of these diagnosis technologies are proposed in this paper. Based on detection technology, a location method for traffic congestion nodes is proposed, which can obtain the health degree of the nodes more clearly. In combination with the advantages of adaptive detection and pre plan detection, a new integrated detection scheme is proposed, which can optimize the cost and time of detection. In addition, a dynamic adjustment strategy is proposed for the detection interval, which is more flexible and reasonable.Existing methods generally divides into the node health and failure two states. Unlike the failed node, congested nodes not completely stops working, it seriously affect the transmission performance of the network. To locate these nodes, and simulate the flow transmission model of the network, a network fault diagnosis model based on flow is proposed, by which the node status are divided into finer. Position detection site selection directly affects the cost and the number of detection station needed for detection. A greedy algorithm is proposed, which can achieve higher monitoring rate in the case of a small amount of nodes are selected as monitoring station. The probe reduction strategy is proposed to reduce the cost of monitoring. And the corresponding probe selection algorithm is proposed. Simulation results show the algorithm can locate the congested nodes and failed nodes accurately and effectively.The pre-planned detection and adaptive detection techniques have their own advantages and disadvantages. The speed of the pre-planned detection is faster, and the adaptive detection’s cost is lower. Combined with the advantages of two kinds of detecting a comprehensive type detection scheme is put forward. Similar to the pre-planned detection, we transmit probe in the two stages of algorithm, according to the idea of adaptive probe we execute the second stage or not is determined by the result of the first stage. The numerical result shows that we can localize failure quickly and the cost of probe is little. We put forward a dynamical adjustment strategy for the time interval of detection, according to health status of network at different time. We can increase interval time so that the cost of probe is little when the network status is better, while we can shorten the time to localize the failure nodes quickly when the network status is worse, make the diagnose more flexible and reasonable.