Research On Energy-saving In CDP Storage Systems

In recent years,with the rapid development of information society,data security has attracted tremendous attention.Meanwhile,Continuous Data Protection(CDP),the technology of data backup,emerges and causes data increasing dramatically,and that makes storage devices increase accordingly.It inevitably leads to a high energy consumption.Under the background of energy-saving and emission-reduction,how to save energy effectively in storage systems is very important for data center.In addition,in the continuous data storage systems,such as CDP storage system,the data has single-feature,written in sequence,without modification,and the load has temporal correlation.In this paper,we propose a storage system,D-RAID,with the dynamic grouping data layout based on characteristics of data storage of CDP.It carries out distribution management dynamically and performance optimization of the RAID storage space,to balance the load demand and energy consumption of the storage system.The storage system could meet the requirements of applications and energy-saving.The specific contents are as follows:1.The theoretical knowledge of the storage technology,including the internal structure,performance indicators and energy consumption characteristics of the storage media,Disk and Solid State drive,is introduced,which are mainstream currently.In addition,a detailed introduction of the typical structure,which appeared during its development of RAID is given and its reliability is analyzed.Finally,the backup system is introduced in the paper.2.The related technologies of CDP and two kinds of data layout structures of fixed grouping are introduced in this paper which analyzes the existed problem of fixed grouping.D-RAID,a dynamic grouping data layout is presented based on that.According to the storage system of CDP,it can adjust the number of the active disks dynamically,making data concentrated in few disks to save energy.In this paper,we propose a optimization strategy for read and write operations,Data Incremental Parity algorithm and prefetching strategy of the old parity data,which can minimize write penalty caused by "read-modify-write",and improve the performance of the storage system.3.The relevant content about the MD module in Linux kernel is analyzed and the energy-saving storage system: D-RAID is implemented in this paper.Moreover,the disk scheduling algorithm and optimization strategy is developed.According to the experimental data of performance and energy consumption test of D-RAID,it is proved that the data layout of dynamic grouping for the load characteristics of real-time variety achieves the desired energy efficiency ratio.