Research On Network Fault Diagnosis Based On Traffic Analyzing

With the increasing complexity and large of the computer network, the frequency of the network failure is also getting higher and higher. Correspondingly, the ability to improve the network fault intelligent diagnosis has become the urgent needs. At present, the network fault diagnosis is still in the primary stage, which not only affect the user experience, but also increased the maintenance costs. Therefore, it is of great significance for service providers to diagnose and locate network faults quickly.This thesis studies how to quickly and efficiently find network failure in the large-scale network traffic, to obtain the preliminary results of fault diagnosis, and then use Bayesian network model to infer the final network failure. The discovery and location of network fault in this thesis is a case study of user login failure. Due to the current network flow is very huge, how to quickly parse network flow and establish a session track record is a problem we need to first resolve. In this paper, the EHT session tracking algorithm is proposed, which greatly improves the efficiency compared to the traditional session tracking algorithm. On the other hand, due to the complexity of the relationship in the network, many intricate coupling, a lot of uncertain factors and uncertain information, the network fault diagnosis becomes more difficult. This thesis will establish Bayesian fault diagnosis model, using Bayesian uncertainty reasoning characteristics and learning characteristics, to do network fault diagnosis quickly and effectively. The main research contents of this thesis include:(1) Research on large scale network session acquisition and tracking technology, put forward the EHT session tracking algorithm.(2) Research on the relevant theory and reasoning of Bayesian networks and Bayesian networks in the study of storage program design expression method.(3) Establish a diagnostic Bayesian network model for fault diagnosis, finally get the fault reason.(4) Design and implement prototype system.In this thesis, the prototype system is tested in the experimental network environment, and its effectiveness is verified.