Research Of Data Disaster Recovery Technologies Based On Hierarchical Model

With the development of information technology, the digital information storage, processing and transmission become important and indispensible components, and the availability and integrity of data are the most important resources of information system. However, the occurrence of disasters is unpredictable and uncontrollable, which may destroy the whole data and computing infrastructure, and lead to the information in the computer damaged or lost. Although the traditional data backup technology and server cluster technology have avoided the damage that is caused by various software and hardware malfunction, human operator error, or viral attacking, with the rapid increase of network data traffic and the application system structures being getting more and more complex, the difficulty of data security protection increases gradually. Therefore, in order to meet the needs of development, we must depend on high-performance data disaster recovery technology, ensuring that the system works normally when a disaster happens.In this paper, first we carry on the thorough analysis and review to the literatures from domestic and foreign related fields, summarize the advantages and disadvantages of current disaster recovery system, and research the key technologies of disaster recovery system by the hierarchical pattern. The paper mainly states the following research work:On the basis of analyzing a variety of disaster recovery system models, we propose a data disaster recovery solution based on the hierarchical model, which designs and implements multi- granularity data disaster recovery technology from different levels of network architecture, stores the data in distributed storage adopting the mechanism of erasure code, employs the replication technology of network layer data stream to achieve incremental data replication, and uses undo technology upon application layer protocol to implement the diversity data replication. We make the methods mentioned above fused in one period of circulating backup, identify the performance indicators of the framework by formal description, and validate the feasibility of the framework through the simulation experiment at last.Based on the research of the hierarchical model, in this paper, firstly we study the full-replication technology to store data, and implement distribution storage of the data by adopting the data redundancy. Through the analysis of multiple redundant codes, we present an encoding technology applicable to the interconnection network application, that is the IRA code storage technology of mirror-oriented mechanism, which distributes the encoded data to distributed storage system and is characterized by taking less storage space, disaster recovery, data integrity, and so on.Full-replication of data can guarantee the security of storage data to a certain degree, however, the data having no time to back up will be lost after a disaster happens. Consequently, we present one data replication technology based on the network layer data stream, which runs the Incremental Backup to the data stream in real-time, and solves the problem of data loss from the full backup point to the moment of error happening. We propose a kind of disaster recovery system architecture based on network layer data stream, and focus on researching the rapid acquisition of data stream and the filter of invalid data packets, the orderly storage technology of the data packets supporting the fast retrieval and the key technique of data streams playback at any point. Finally, experimental results on testing the performance of the system demonstrate that the proposed method has little effect on the application system operating normally.Replication technology based on data stream occupies large storage space as a result of adopting real-time data replication. In the middle of the period of data stream replication, inserting the backup of diversity data can solve the problem of occupying large storage space. In the paper we put forward a diversity data replication technology upon the action predicate of application layer. And the replication method records the actions after full-replication, only playbacks the data changed due to the action of application layer when data is recovered, hence it can greatly save storage space, but also improve the situation of packet loss under high flow. Although the method relays on the specific protocol of application layer, its action of fusion and undo method improve the recovery ability of the system. Additionally, we design and implement the undo mechanism based on 4R model, and comparative analysis validates the feasibility of the system by experiment.In this paper, through the hierarchical approach we discuss the data disaster recovery system in detail, implement a variety of granular data recovery techniques, and guarantee the security of stored data. The research of this paper will have an important research significance and application value on the development of data disaster recovery.