| In the IT application scenarios,data security and availability are the basis of the entire application system.Special data protection strategy is needed for the key data.Continuous data protection technology can provide fast,arbitrary fine granularity data recovery capability.This paper focuses on the key technologies for data storage management oriented continuous data protection.A file level continuous data protection based on distributed object storage is proposed;and the technology of deduplication is studied to deal with the redundant backup problems caused by continuous data protection.The main contributions of this paper include the following aspects.1)This paper proposes a file level continuous data protection OB-CDP.In this paper,the metadata structure,management,backup method and data object storage scheme of OB-CDP are designed,a file system driver based on FUSE framework is implemented.A hash index based metadata management method and a controllable granularity backup time window method are proposed to optimize OB-CDP write delay.Tests show write delay can reduce 17%.2)This paper theoretically explains the reason for low throughput and large chunk size variance problems in rabin fingerprint based chunking method.A method of skipping fixed length chunk is proposed to accelerate the chunking rate and a two chunking thresholds method is proposed to normalize the chunk size variance.Tests show that proposed variable chunking method accelates chunking rate by 58%,more than 90% chunks are distributed near the expected chunk size.3)This paper proposes a fingerprint locality based fingerprint index method using the segment structure to preserve locality in the backup data fingerprint.And a segment caching and prefetching method is proposed to further improve the fingerprint index performance.4)This paper designs a prototype system for OB-CDP.This paper also implements the chunking method and index method the on the open source deduplication platform. |
PDF Full Text Request