The Research Of Disk Array Reliability

With the explosive growth of digital information, multimedia applications and theincrease of numbers of user groups, the demands for mass information storage systemsalso have grown. Large amounts of data also take a challenge to the data storage system,especially for the customers like mobile operators and the banking and financial system,who have higher reliability requirements for data storage systems. At today’s storagesystems, disk is the essential device to constitute mass storage system. Therefore, howto find the Latent Sector Error and repair it quickly, and construct a disk storage systemwith high reliability, high performance, become an important task.At present, there are mainly two major categories of methods used to eliminate theinfluence of Latent Sector Errors to data storage system. One method is Intra-DiskRedundancy Scheme, such as the Interleaved Parity-Check(IPC) Coding andReed-Solomon(RS) Coding. By increasing the data redundancy, they reduce the impactof Latent Sector Errors on the disk reliability. The other method is Disk ScrubbingScheme, whose most important role is to detect and repair the Latent Sector Errors ondisk as soon as possible, by sending orders to read IO from disk, therefore improves thereliability of the disk array. In this paper, we present a new Scrubbing Scheme, whichwe call smart scrubbing scheme, which is put forward on the basis of the deterministicscrubbing scheme and random scrubbing scheme. Combining the advantages of bothdeterministic scrubbing scheme and random scrubbing scheme, smart scrubbing schemecan find the Latent Sector Errors more quickly, and improve the reliability of the datastorage system.With the growth of data, the appearance probability of Latent Sector Error in thedata storage system has also increased. However, Interleaved Parity-Check(IPC) Coding,Reed-Solomon(RS) Coding, deterministic scrubbing scheme and random scrubbingscheme are only algorithms which provide preventive measures, they can fix thediscovered single disk bad track, but will be out of action and result in disk array failureand data loss when occurred during the data reconstruction process of disk array,especially when disk array redundancy has already reached the upper limit and otherdata disks show one or more disk bad tracks. In this paper, we present a new technique, which call bad block marker technology. It can mark the disk sector where the disk badblocks exist, recover some other data on the failed disk through reconstruction, andmeanwhile the disk array is still available, thus improves the reliability of the disk arrayand minimize the user’s data loss and its impact.