Research And Implementation Of Intelligent State Early-Waring And Backup System

Disaster backup is very important for the business data security.In order to prevent and resist the losses caused by natural disasters and man-made disasters,it is necessary for important business systems and data to do best disaster backup measures.The traditional disaster recovery technology determines that the data center based backup of the traditional information system has been realized in a passive mode,and its operation and maintenance human cost and risk can not meet the needs of the current information system.With the improvement of real-time business requirements for system service continuity and business-level availability,how to intelligently backup data from the existing behavior mode,or effectively predict the disaster,so as to make data backup more efficient and ensure business continuity has become a new research direction of disaster recovery technology.To meet the business needs and development trend of information system disaster recovery technology,this paper integrates the current deep learning technology,knowledge graph,cloud computing virtualization technology and other related technologies,and carried out deep research about the time series data of information system operation,and then modeling the multi-dimensional analysis based on these results,realized the intelligent early warning of the information system state,and carry out a certain degree of evaluation and prediction of the disaster with the proactive disaster recovery mode.At the same time,in the aspect of backup and recovery,virtualization technology is used to deploy the backup data and system applications to the cloud platform through transformation,so as to improve the business availability and continuity.The main research work of this paper is as follows:(1)With the poor generality,robustness and practicability of existing anomaly detection methods based on log data,this paper proposes a multi-level weighted anomaly detection mechanism based on system log data,which optimizes the above problems and solves problems of specific scenario,such as unstable log data(update for example).In order to improve the robustness,this paper uses word-level weights for sentence embedding vectorization and LSTM deep learning network for modeling.At the same time,sentence-level weights is used to explore the interdependence between different log sequences,and a workflow mechanism is constructed to find the abnormal points in the execution process of a complete task flow,so as to improve the accuracy of proposed model.The practical results show that the multi-level weights mechanism can effectively improve the accuracy and reduce the complexity.(2)Aiming at the problems of high labor cost and single detection rules in the existing abnormal diagnosis methods in operation and maintenance work,and based on artificial intelligence for IT operations(AIops),this paper proposes an anomaly root cause recognition mechanism which combines knowledge graph and causal graph mining technology.It can identify and locate anomalies in the distributed dynamic cloud environment with complex interaction.The specific implementation is to obtain directed acyclic graph(DAG)from the operation and maintenance monitoring metrics data by using FCI(fast routine inference)algorithm.At the same time,domain knowledge is used to construct knowledge graph,which simplifies calculation and improves efficiency of FCI algorithm.This paper deployed the micro-service architecture application in the cloud environment and verified the simulated abnormal data.The practice results show that the above abnormal root cause mechanism can effectively explore the abnormal root cause in the cloud computing micro-service architecture,and provide feasible basic support for intelligent disaster recovery.(3)This paper uses virtualization technology to develop and build an intelligent disaster recovery management platform,which provides fine-grained backup services with whole machine backup,application backup,file backup,block-level backup and other aspects.Most importantly,the research work deploys the backup data system application to the cloud platform through transformation,so as to improve the business availability and continuity.