Research On Network Fault Diagnosis Algorithm Based On Data Analysis

Network equipment failure is inevitable,thus effective fault diagnosis technology and research of identification of critical nodes in a large number of associated faults both are important for keeping reliable network communication.To solve fault diagnosis,one of the problems is how to obtain the real-time status information of network equipment from the network with dynamically updated environment(topology updating,device state dynamic changing and uncertain detective data,etc.).On the other hand,mining effective fault information from massive data(detective data,device status information and topology data,etc.)is meaningful for inferring the real faulty device.Large-scale cascading failures are unavoidable because the equipment connection relationship requires reliable critical node identification technology to identify the critical node from a large number of failure nodes.Detailed researches are described as follows:1.This research analyzes the main problem and advantages and disadvantages of diagnosis modes in different research stages firstly,and then explores challenges that fault diagnosis faced in existing large-scale networks.To solve the large-scale cascading-failure in the network,the technology of identifying critical nodes from a large amount of fault nodes is analyzed,and the corresponding problems are to be further explored,which also is described in detail.2.In the terms of obtaining detection data,the proactive fault diagnosis is in good performance for getting real-time status information of network devices.In the traditional mode,the detection stations are arranged in the network and then probes are sent to obtain the state of equipment,which is not suitable for the network with dynamical updating topology and device status.Based on the system-level network fault diagnosis theory and the characteristics(high flatness and high connectivity)of new generation networks,this study designs the heuristic breadth-first fault detection mode(HBFD),this mode is suitable for network with dynamical topology and uncertain detection data.3.The real-time change of network elements status,dynamical business trend and the loss of detection data during transmission make fault reasoning in a dynamic environment get more and more important.This research introduces the Dempster-Shafer dynamical combination theory to establish the dynamical fault reasoning model.Combined with the component of fault detection,a heuristic fault diagnosis model is designed,which employs the dynamic spanning tree searching(DSTS),the fault probability evaluation(FPE)and the fault reasoning(FR).4.To solve the cascading failures in large-scale networks,this research introduces link betweenness centrality to establish a reasonable load redistribution model on the basis of Scale-Free complex network model.Further through simulating the process of cascading failures,the data of nodes topological attributes and evaluation of cascading failure is collected respectively.Combined with the technology of correlation analysis and clustering analysis,the critical node can be identified effectively in large-scale network cascading failures.