|
RAID systems were proposed in the late 80s as a way to use parallelism
in a multitude of disks to improve the aggregate I/O performance.
Today, RAID systems are an integral part of the product lines of
most major computer and storage systems manufacturers. RAID systems
rely on data striping across multiple disks to improve performance
and on redundancy to improve reliability. For large disk arrays,
reliability is a major concern. As the storage capacity and the
data rate of each disk in an array increase, the traditional methods
of mirroring or of simple, single-parity coding are no longer sufficient.
In this project, the ZRL team is exploring advanced coding and
performance optimization techniques to improve resilience to various
kinds of failure modes encountered in modern RAID systems.
Disk arrays are actively used today as a means to improve aggregate
input/output (I/O) performance of storage systems. Modern disk-array-based
data storage systems employ two main architectural techniques, namely
(1) data striping across disks to improve performance and (2) data
redundancy to improve reliability.
Although state-of-the-art disks are highly reliable devices, they
occasionally fail to read or write certain sectors of data. Currently,
most hard-disk drive manufacturers guarantee an uncorrectable bit
error rate of  to
 .
The problem of such uncorrectable errors is more pronounced today
because of the increased capacity and data rates of the disk drives.
This challenge, coupled with disk-drive failures, makes existing
RAID systems vulnerable to data loss. Our group is exploring advanced
coding and performance optimization techniques for improving resilience
to various kinds of failure modes encountered in RAID systems.
References
| [1] |
Ajay Dholakia, Evangelos Eleftheriou, Xiao–Yu
Hu,
Ilias Iliadis, Jai Menon, KK Rao, Analysis of a New Intra-Disk
Redundancy Scheme for High-Reliability RAID Storage Systems
in the Presence of
Unrecoverable Errors, SIGMETRICS 2006, Research Report RZ3652. |
|
