Research On Key Technologies Of Data Center Physical Infrastructure Monitoring System

Data center is a platform for data collection,storage,sharing and processing.As the physical environment for information system operation,the stability,reliability and security of data center infrastructure determine the continuity of it's business.To ensure a good user experience,it is of great significance to maintain the safe and stable operation of the data center.The monitoring technology of data center infrastructure is the foundation of data center operation and maintenance.Finally,the monitoring system will send the alarm or fault information to the corresponding operation and maintenance personnel through SMS,e-mail,etc.,which will be solved and processed by the operation and maintenance personnel to complete the maintenance of the data center infrastructure.At present,the monitoring items in the infrastructure monitoring system of the data center are generally alarmed by setting the threshold value through the manual experience,this method can not describe the fluctuation of the monitoring item data,which may cause the missing alarm and error alarm;Due to the variety of monitoring projects,many monitoring items may bring alarm storm,it is difficult for the operation and maintenance personnel to check and process one by one,so this thesis mainly studies the alarm technology of data center infrastructure monitoring,and puts forward the corresponding solutions to the above problems.The main work of this thesis is as follows:Firstly,this thesis studies the related technologies of the data center infrastructure monitoring system,combs the workflow of the monitoring system,and focuses on the alarm technology of the monitoring system.Secondly,for the traditional threshold alarm method in the monitoring system,this thesis proposes a no-threshold alarm method based on EFT(Energy;Fluctuation;Time),By considering the characteristics of each monitoring item in data center infrastructure,this method proposes to extract features from three aspects of monitoring item data: energy,fluctuation and time.And the classification algorithm inmachine learning is introduced into the alarm field of monitoring items,and finally the no-threshold alarm model based on EFT is obtained to realize the no-threshold alarm of monitoring items,The experimental results show that the accuracy of the proposed no-threshold alarm method through random forest classification is over95%.Finally,due to a large number of data center infrastructure monitoring projects,many monitoring items alarm,the workload of operation and maintenance personnel troubleshooting one by one is large,In order to further reduce the workload of operation and maintenance personnel,this thesis proposes a fusion analysis method of monitoring item alarms,Through the root cause analysis between the alarm results of the monitoring items,the fusion alarm results of each monitoring item are obtained,and the fusion analysis model of the monitoring items is obtained through the training model of the classification algorithm.The alarm fusion analysis method proposed in this thesis can reduce the alarm amount of monitoring items and reduce the work pressure of operation and maintenance personnel.