Characterizing Declustered Software RAID for Enhancing Storage Reliability and Performance

Authors: Zhi Qiao (University of North Texas), Song Fu (University of North Texas), Hsing-bung Chen (Los Alamos National Laboratory), Bradley Settlemyer (Los Alamos National Laboratory)

Abstract: Redundant array of independent disks (RAID) has been widely used to address the reliability issue in storage systems. As the scale of modern storage systems continues growing, disk failure becomes the norm. With ever-increasing disk capacity, RAID recovery based on disk rebuild becomes more costly, which causes significant performance degradation and even unavailability of storage systems. Declustered parity and data placement in RAID aims to enhance the recovery performance by shuffling data among all disks in a RAID group. All disks in the RAID group participate in data reconstruction, which leads to reduction of the RAID rebuild time. In this work, we extensively evaluate declustered RAID in terms of the performance of application I/O and recovery time. Our experimental results in ZFS show that the speedup of declustered RAID over traditional RAID is sub-linear to the number of disks in the storage pool.

Best Poster Finalist (BP): no

Poster: pdf
Poster summary: PDF
Reproducibility Description Appendix: PDF


Back to Poster Archive Listing