Research And Implementation Of Continous Data Protection System Based On Block Level

With the development of Internet, the amount of data needed to be stored is increasing. The instability of hardware and software may lead to the bss of user data So the reliability of data storage has become a hot topic in industrial and academic area.The reliability of data storage could be guaranteed in the hardware and software. In the hardware, RAID (Redundant Arrays of Independent Disks) is used to solve the corresponding problems. In the software, the traditional technology cannot meet the demand of the reliability. It recovers the data by historical versions. But the existence of backup window cannot prevent the loss of data effectively. However, CDP (Continuous Data Protection) is a new technology which can eliminate backup window completely. When the system loses data due to software errors, CDP could restore the data to the state user’s desire quickly and correctly. The main content and achievements of this paper are as follows:(1)This paper reviewed various data protection technologies and analyzed their principle, advantages and disadvantages. By experiments, we then selected IP-SAN architecture as the solution for data protection.(2)We made a detailed study and analysis of the key technologies of CDP, which includes data capturing and restoring of write operations. Meantime, we researched the recovery strategies of the data Trough experiments, we evaluated the performance and verify the validity of CDP.(3)A Mirror and Snapshot CDP (MSCDP) was designed and implemented. This system was used to deal with software errors on small and medium server. It was a combination of technologies that includes snapshot, mirror and CDP on IP-SAN. Using ISCSI protocol for communication between application servers and storage servers, the data can be restored without loss when errors occur. MSCDP designed the data structure of the disk, which was used to store the captured data, and adopted a hierarchical process to conduct the read and write operations. We used BIO directly to accelerate the speed. In the recovery phase, MSCDP decreased the time of recovery by optimizing index structure.