Research On High Productivity Of Continuous Data Storage System

In the era of big data,storage devices are doubled every 18 months with data increasing dramatically.That cause to a problem which cannot be ignorable for data centers: high energy consumption.Storage systems with continuous data access,such as,video surveillance,data backup and archiving storage,need not high data transfer rate,and data in it is written once,not modify,and read maybe.Power consumption and performance are deeply studied based on the workload characteristics of continuous data access in this paper.1)The current situation and main medium of storage system are introduced and analysed.The performance and energy consumption of Disks and SSDs are described and compared.7 major RAID structures(RAID 0~RAID 6)are presented in the process of RAID development.2)According to traditional RAID,we propose an alternative RAID data layout: SHG-RAID(Horizontal Group of Stripes in RAID).In SHG-RAID,data blocks are grouped,and data is written into data blocks group by group.In a period of time,there only one group disks are written into,and then,the other disks in storage system could be turn into standby mode for energy-saving.Turning disk into standby mode also degrades the performance of the storage system,and setting different numbers of disks in one group when the system initializing can suit for different performance requirements.Experiments show that,SHG-RAID with 12 disks could save 49% power consumption than traditional RAID.SHG-RAID can save energy dramatically at the cost of lowering data transfer speed but can still meet the applications request comparing to traditional RAID.3)Based on the characteristics of HS-RAID data layout,we propose a power-aware RAID data layout: HS-RAID2(HS-RAID with DIP),saves energy by grouping disks in the array,meanwhile,radom data requests are redirected to SSD to optimize energy conservation.However,all of the write operations in HS-RAID are small write,and they degrade the storage system's performance severely.A redundancy algorithm is proposed: data incremental parity algorithm(DIP),applied to HS-RAID to minimize the write penalty and improved the performance and reliability of storage systems.Results show that HS-RAID2 with 2 SSDs and 12 disks has at least 52.3% speedup and higher reliability than HS-RAID.