Data redundancy is necessary to prevent instances
in which a single disk failure can cause a system to go down until
the problem is located and corrected. There are two methods of providing
data redundancy: software mirroring and hardware mirroring. Each represents
RAID Level 1. (See “Using Disk Arrays” for more information on the meaning
of the various RAID levels.)
Software mirroring allows you to maintain identical copies of your data (except for
the root disk), so that each set of data has, in effect, a perfect
clone of itself. In the event a disk fails, the system can use the
mirrored copy of the data, thus allowing users to continue to work
without interruption. The bad disk can be replaced at a more convenient
time when the system can be brought down without causing problems.
Once the system is rebooted, the mirroring software will cause the
mirrored data to be copied back to the replacement disk and the process
of mirroring will begin again.
With three-way disk mirroring, two copies of each disk’s data are maintained. This strategy
is even more robust than two-way mirroring which is described above
and it eliminates the need to bring the system down at all in order
to replace a bad disk.
|
| |
|
| NOTE: Using Version 1 LVM Volume Groups, you can have up to three
copies of your data (the original plus two mirror copies). Using Version
2 LVM Volume Groups you can have up to six copies of your data (the
original plus five mirror copies). |
|
| |
|
To use disk mirroring,
you will need to use LVM or VxVM as your disk management strategy
and (if you are using LVM) have available the MirrorDisk/UX software
product. MirrorDisk/UX causes every write to the original volume to
also be written to the copy or copies of the original volume. Note
that the original data and its copied data may be spread over more
than one disk.
The main advantage of software mirroring over
hardware mirroring, which is discussed in “Using Disk Arrays”, is that the cost of implementation
is lower. The main disadvantage of software mirroring relates to its
increased complexity of management. That is, it will probably be significantly
more difficult to manage a system with a large number of disks as
compared to a system with a single disk array.