The Design Of Snapshot And CDP Based On Storage Virtualization

As the mode of scientific computing is migrating from the traditionalcompute-intensive computing to data-intensive computing, data gets to becomeincreasingly important for the operating of enterprises or institutions. Even a slightamount of data unavailability may lead to incalculable losing. Traditional dataprotection technologies usually focus on how to protection the data of some specificstorage nodes or storage applications. But they are incapable to deal with the newproblems arising in the process that storage system is getting large-sale, networkingand virtualization. The organization and structure of storage system is far morecomplex than ever before, and the applications running on storage system are varygreatly in their types and requirements. This thesis implemented high performancesnapshot system and high utility CDP system at logical volume level based on storagevirtualization technology. The two systems not only adapt to different underlyingphysic storage devices of large-scale storage system, but also provide data protectionservice for many kinds of upper layer applications. The main works of this thesis areas follows:1. Proposes A Logical Volume High performance snapshot (LVHPsnap) method.With a scalable metadata managing strategy, the storage system can freely supportmore than one type of snapshot volume to meet different requirements of applications.By compacting snapshot index data and constructing composite snapshot volume listwith Finesnap volume and Checkpoint volume, storage system maintains highperformance and high resource utilization ratio while time to recovery from snapshotvolume being short. The snapshot system LVHPsnap built on technologies aboveperforms good performance in representative experiments.2. Adopts A self-adaptive snapshot generation policy. Then, modifications to datablock will be traced, and the time interval of generating a new snapshot volume can bedynamically shifted to follow the waving of application loads. So the historical datacan be automated stored when needed, and the system can also maintain a high levelof performance and resource utility. This self-adaptive snapshot generation policygains the advantages on its load waving adaptability, easy implementing, and havingno impact on I/O performance.3. Proposes A high utility and flexible CDP method. The method is flexible to realize among the storage nodes in the storage virtualization environment, and itsimpact on I/O performance is slight. By implementing the CDP method, data can beconveniently retrieved to the state of any point of time. Also, an of out-of-band copymechanism is adopted to backup the original data, and the I/O service doesn’t need tobe suspended any more during the process of CDP initiation.