Fast Recovery and Low Cost Coexist: When Continuous Data Protection Meets the Cloud

Yu GU  Chuanyi LIU  Dongsheng WANG  

IEICE TRANSACTIONS on Information and Systems   Vol.E97-D   No.7   pp.1700-1708
Publication Date: 2014/07/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E97.D.1700
Type of Manuscript: Special Section PAPER (Special Section on Cloud and Services Computing)
CDP,  block-level,  cloud,  deduplication,  MYCDP,  

Full Text: PDF>>
Buy this Article

Cloud computing has rising as a new popular service paradigm with typical advantages as ease of use, unlimited resources and pay-as-you-go pricing model. Cloud resources are more flexible and cost-effective than private or colocation resources thus more suitable for storing the outdated backup data that are infrequently accessed by continuous data protection (CDP) systems. However, the cloud achieves low cost at the same time may slow down the recovery procedure due to its low bandwidth and high latency. In this paper, a novel block-level CDP system architecture: MYCDP is proposed to utilize cloud resources as the back-end storage. Unlike traditional delta-encoding based CDP approaches which should traverse all the dependent versions and decode the recovery point, MYCDP adopts data deduplication mechanism to eliminate data redundancy between all versions of all blocks, and constructs a version index for all versions of the protected storage, thus it can use a query-and-fetch process to recover version data. And with a specific version index data structure and a disk/memory hybrid cache module, MYCDP reduces the storage space consumption and data transfer between local and cloud. It also supports deletion of arbitrary versions without risk of invalidating some other versions. Experimental results demonstrate that MYCDP can achieve much lower cost than traditional local based CDP approaches, while remaining almost the same recovery speed with the local based deduplication approach for most recovery cases. Furthermore, MYCDP can obtain both faster recovery and lower cost than cloud based delta-encoding CDP approaches for any recovery points. And MYCDP gets more profits while protecting multiple systems together.