Research On Capacity Promotion Technology Of Disk Array And Disk In Block Level

With the explosive growth of data volumes, especially with the rapid development of big data applications in recent years, data storage system faces severe challenges, including the heavy pressures on storage capacity and performance. Moreover, continued growth in the amount of data needs the continued expansion of storage capacity. Thus, storage devices should have the capability of online capacity expansion and this forms an important research orientation.As a wide applicated storage device, disk array must fit the character of continuous increasing of data to have the ability of online capacity expansion and performance enhancement after expansion. To expand the capacity of a RAID-5array with additional disks, data need to be migrated between disks in order to leverage extra space and performance gain. Conventional methods for expanding RAID-5are very slow because they just have to migrate almost all existing data and recalculate all parity blocks. This dissertation presents a new online capacity expansion method for RAID-5based on parity blocks migration (PBM). This method only migrates blocks that form a special parallelogram on data topology graph with one side consisting of only parity blocks. When adding m disks to a RAID-5with n disks, PBM achieves the minimal data migration which only needs to move m/(n+m) of all data blocks. Furthermore, no parity blocks are recalculated during the expansion. After expansion, although the RAID is not a standard RAID-5distribution, the parity blocks are distributed evenly. We do experiments in real disk array with Linux. Experimental results show that, on average, PBM can reduce the time of expansion by73.6%while only reduces the performance of the expanded RAID by1.83%when compared with Multiple-Device Reshape(MD).The dissertation proposes a novel expansion method HCS which aimed at a RAID6coding, H-Code. H-Code is a vertical coding, and it has good expandable personality. HCS achieve no parity blocks recalculation, including P and Q, with anti-diagonal data selection and vertical data migration. After expansion, the disk array can keep data distribution as H-Code manner, thus HCS can achieve multiple expansion. Experimental results show that HCS can decrease expansion time about3.6%and promote performance about4%, compared with SDM which is another RAID6expansion method.Disk as the foundational storage device, several technologies can expand its capacity, and the shingled recording disk is a typical one. Shingled recording disk can achieve over1Tb/in2storage density with overlapping some tracks. This dissertation proposes a new wave-like track layout based traditional shingled recording track layout. Wave-like track layout can promote storage density about6%-10%with bi-direction track overlapping. For improving the performance of wave-like shingled recording disk, the dissertation studies a hybrid shingled recording storage system, HWSR. HWSR divide the shingled recording disk into segments area. It uses solid state disk (SSD) buffer to reside read requests data and uses memory buffer to reside write requests data. A new data eviction algorithm based on maximum write efficiency HWSR-ER is used when data write to shingled recording disk from memory buffer, thus the performance of HWSR is promoted. Experimental results under8I/O workloads show that this hybrid system can upgrade average response time of38.2%compared with normal disk.