Software RAID considerations
The Linux kernel software RAID driver (called md, for multiple device) offers integrated software RAID without
the need for additional hardware disk controllers or kernel patches. Unlike most hardware RAID solutions,
software RAID can be used with all types of disk technologies, including SATA, SCSI, and solid-state drives.
This software solution requires only minimal setup of the disks themselves.
However, when compared to hardware-based RAID, software RAID has disadvantages in managing the disks,
breaking up data as necessary, and managing parity data. The CPU must assume some extra loading: disk-
intensive workloads result in roughly double the CPU overhead (for example, from 15% to 30%). For most
applications, this overhead is easily handled by excess headroom in the processors. But for some applications
where disk and CPU performance are very well balanced and already near bottleneck levels, this additional
CPU overhead can become troublesome.
Hardware RAID offers advantages because of its large hardware cache and the capability for better
scheduling of operations in parallel. However, software RAID offers more flexibility for disk and disk controller
setup. Additionally, hardware RAID requires that a failed RAID controller must be replaced with an identical
model to avoid data loss, whereas software RAID imposes no such requirements.
Some software RAID schemes offer data protection through mirroring (copying the data to multiple disks in
case one disk fails) or parity data (checksums that allow error detection and limited rebuilding of data in case
of a failure). For all software RAID solutions on HP workstations, redundancy can be restored only after the
system is shut down so that the failed drive can be replaced. This replacement requires only a minimum
amount of work.
Performance considerations
Disk I/O bandwidth is typically limited by the system bus speeds, the disk controller, and the disks
themselves. The balance of these hardware limitations, as affected by the software configuration, determines
the location of the any bottleneck is in the system.
Several RAID levels offer improved performance relative to stand-alone disk performance. If disk throughput
is restricted because of a single disk controller, RAID can probably do little to improve performance until
another controller is added. Conversely, if raw disk performance is the bottleneck, a tuned software RAID
solution can dramatically improve the throughput. The slower disk performance is, relative to the rest of the
system, the better RAID performance will scale, because the slowest piece of the performance pipeline is
being directly addressed by moving to RAID.
Configuring software RAID
See the following sites for additional information about configuring software RAID on Red Hat Enterprise
Linux (RHEL) or SUSE Linux Enterprise Desktop (SLED):
●
Red Hat Enterprise Linux 6—See the Storage Administration Guide at
https://access.redhat.com/site/
documentation
●
SLED 11—See the Deployment Guide at
http://www.suse.com/documentation/sled11/
book_sle_deployment/?page=/documentation/sled11/book_sle_deployment/data/
book_sle_deployment.html
.
For more information about Open Source and Linux at HP, go to
http://www.hp.com/go/opensource
.
82
Appendix B Configuring RAID devices
