Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6858648 | Information Systems | 2015 | 15 Pages |
Abstract
Software redundant arrays of independent disks (RAID) suffer from several hours of resynchronization time after a sudden power-off. Data blocks and a parity block in a stripe must be updated in a consistent manner. However, a data block may be updated without a parity update if power goes off. Such a partially modified stripe must be updated with a correct parity block after a reboot. It is difficult, however, to find which stripe is partially updated. The widely used traditional parity resynchronization approach entails a very long process that scans the entire volume to find and fix partially updated stripes. As a remedy to this problem, this paper presents a parity resynchronization scheme that exhibits a small overhead for a wide range of workloads, finishes parity resynchronization within several minutes, and is transparent to file systems, thanks to a new seamless block-level journaling. The proposed scheme is integrated into a software RAID driver in a Linux system. A performance evaluation demonstrates that the proposed scheme shortens the resynchronization process from 200Â min to 30Â s with 1% overhead, compared to 51% overhead for the prior scheme.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
Sung Hoon Baek, Ki-Woong Park,