Effects of Data Scrubbing on Reliability in Storage Systems

Today, even bytes stored in expensive, professionally managed data centers, are vulnerable to loss and damage. The vulnerability increases when large volume of data must be stored indefinitely into the future, as required by emerging data services. The economic viability of these services depends on their keeping data unaltered. As disks outpace tapes in capacity growth, large scale storage systems based on disks are becoming increasingly attractive. Such systems will encompass a very large number of disks and store petabyte of data. In large systems encompassing thousands or even tens of thousands of active disks, disk failure will become frequent, necessitating redundant data storage. Until now, the reliability measure in disk drive-based storage systems focus predominantly on protecting against complete disk failures. However, very little is known about the incidence of latent sector errors i.e., errors that go undetected until the corresponding disks sectors are accessed. Hence, periodically accessing the entire storage (data scrubbing) is necessary to protect data loss due to latent sector errors. In this research, we made a standard Markov failure and repair model with data scrubbing mechanism to measure the reliability of the storage system. And we determined the replicas and the period of the data scrubbing for the demanded reliability of the disk-based storage system by using the proposed Markov model.