Study On Technology Of Data Error Tolerance And Disaster Tolerance In Network Storage

With the rapid development of information technology, the application mode of computer system evolves from computing to data processing. As a result, the importance of computer system lies more in processing and storage of data. So the value and importance of data are usually more important than that of the computer system equipment. The data error tolerance and date disaster tolerance technology are effective to improve the data availability, so the research on them has the important theory significance and application value.This thesis is supported by 2003 Aviation Science Fund Project under Grant No. 03F53031 and Xian Industry Plan Project under Grant No. GG200312. This thesis studies the data error tolerance and date disaster tolerance technology under network storage architecture. A storage virtualization model based on iSCSI protocol is presented, then the security transmission scheme, multi-error tolerance data placement algorithm, heterogeneous storage system data placement algorithm, and data disaster tolerance model are researched in detail.The main work and the innovation of the study are as follows.1. While iSCSI protocol does not consider the secure transmission of data, it is vulnerable to malicious attacks and can potentially incur the compromise of the integrity and confidentiality of storage data. The iSCSI security scheme based on IPSec and the iSCSI security scheme based on SSH are presented, which protect the secure transmission of data with lower system overhead by performing secure processing in different layers of network protocol.2. The single-error tolerance data placement can not satisfy application requirement to the data availability. Multi-error tolerance data placement algorithm DP-RAID and TP-RAID are proposed, which are based on horizontal parity and diagonal parity. DP-RAID can tolerate two simultaneous disk errors, and TP-RAID can tolerate three simultaneous disk errors. Compared with similar data placement, the encoding and decoding of our placement is simpler, computing load is lighter, and the performance of small write is better.3. Current data placement fails to consider the capacity and I/O performance difference of different disks. A heterogeneous storage system data placement algorithm PA-RAID0 is presented. PA-RAID0 partitions placement pattern according to capacity difference of different disks, and arranges address block in placement pattern according to I/O performance difference of different disks. This placement achieves 100ï¼…storage capacity utilization, and makes full use of the I/O performance of all disks, which improves the concurrency of the storage system.4. The main problem of the asynchronous data replication is that it can not ensure the data consistency between primary and secondary storage systems. An asynchronous remote data replication protocol ARPDL is presented. ARPDL decreases the network traffic by using "write coalescing" and "asynchronous propagation", and avoids the inconsistency of primary and secondary storage systems by an atomic update scheme. ARPDL ensures that storage system can rapidly respond to I/O request, and ensures data consistency between primary and secondary storage systems.5. General data disaster tolerance technology based on remote data replication maintains one or several pieces of data redundancy in different places, and achieves higher data availability, but it is only effective for the environment with single data nodes. As to the environment with many data nodes, it can not achieve higher data availability by maintaining different places redundancy for each data node in addition to higher storage overhead. A data disaster tolerance model based on erasure coding is presented, which aims at the environment with many data nodes. The model can achieve excellent data availability with lower storage overhead.