Design And Implementation Of RDP Algorithm For Tolerating Two Disk Failures

With the development of storage technology, the reliability of data storage is more and more emphasized. RAID is much popular in all kinds of the users, because it can not only provide high reliability, but also achieve high bandwidth and huge capacity. RAID has different level, which can satisfy different need of reliability. At present, the mellow tolerating disk failure technology is the level of five, which can tolerate one disk failure. Its reliability is not enough.The technology of Tolerating two disk Failures based on the strict mathematics theory, can provide data-access currently even if two of disks miss data by depositing redundant parity data. This level of RAID can be implemented by much algorithm, and each has its strong point, however, we have to choose one algorithm which can not only keep universal, but also have high performance. So is RDP (Row Diagonal Parity). RDP, as a optimal algorithm for two disks failure correction (RAID6), can improve the performance of RAID6 efficiently, but the disks of composing RAID6 is limited by the algorithm, which make capacity wasted, and not convenient for function addition such as online capacity expansion and online level change.In order to overcome the limited of disks, dummy some member disks of RAID6 is proposed. According to implementing single stripe RDP and this algorithm, the performance is improved and at the same time the limitation is released. By the real test, comparing with Reed-Solomon algorithm in the similar condition, the performance of this algorithm has improved 23.1 percent in sequence-write and 18.3 percent in sequence-read.What's more, in order to make this disk array self-contained, Online Capacity Expansion is implemented on this RAID6. Online Capacity Expansion technology is a very useful function in RAID, which can bind new disks into running RAID6. When the capacity expansion is running, which makes RAID6 using much facility.