Research And Application Of Root Cause Analysis Of Abnormal Behavior In Ethernet Network

As the carrier of data transmission,network will inevitably appear abnormal situations in the process of operation.It is impossible to timely understand the operation status of equipment in the network environment only by human resources.Once there is an anomaly,if it is not found and solved in time,it will inevitably affect the efficiency of data transmission,even lead to network paralysis,and affect the production environment.Network state monitoring technology can be divided into active monitoring and passive monitoring.Active monitoring technology will affect the performance of network environment and equipment because its measurement nodes inject sample data into the network environment.Passive surveillance technology measures the network operation status by setting threshold parameters artificially,which is affected by personal experience and can not accurately measure whether it is abnormal.And none of them has the function of deep analysis of abnormal data.Based on the advantages and disadvantages of active and passive network monitoring technology,this paper takes the network delay,packet loss,traffic,CPU status,memory status and log information as data objects on the basis of passive network monitoring technology.Analysis of data characteristics,anomaly data detection,anomaly root cause analysis and other functions.The function of anomaly data detection combines Kmeans algorithm,normal distribution algorithm and dynamic threshold method to solve the problem of inaccuracy of manual intervention in traditional network monitoring.The proportion of abnormal data detected by the system is not more than 0.2%,which is in line with the theoretical expectation of the algorithm;and the proportion of abnormal data is higher than 13.5% in the way of setting threshold parameters artificially.Obviously,the function of abnormal data detection in this system is more in line with the actual situation.The root cause analysis function of abnormal data simulates the thinking logic and method when dealing with network abnormalities manually,and analyses the causes of the abnormal data.The accuracy of the test is as high as 88%.The experimental results show that the system has better data processing performance and accuracy in network anomaly detection and root cause analysis,and can shorten the time consumed by operation and maintenance personnel in dealing with network anomaly and data collection time.There are 56 papers,22 tables and 45 references.