With virtualization, ensure that the storage devices are
configured
to provide some type of redundancy against hard disk failures.
A failure of a storage device can affect a larger
amount of storage
that is presented to the hosts. To provide redundancy, storage devices
can
be configured as RAID arrays that use either mirroring or parity to
protect
against single failures.
When creating RAID arrays with parity
protection
(for example, RAID-5 arrays) consider how many component disks you
want to
use in each array. If you use a large amount of disks, you can reduce
the
number of disks that are required to provide availability for the
same total
capacity (1 per array). However, more disks mean that it takes a longer
time
to rebuild a replacement disk after a disk failure, and during this
period
a second disk failure causes a loss of all array data. More data is
affected
by a disk failure for a larger number of member disks because performance
is reduced while you rebuild onto a hot spare (a redundant disk) and
more
data is exposed if a second disk fails before the rebuild operation
is complete.
The smaller the number of disks, the more likely it is that write
operations
span an entire stripe (
stripe size, multiplied
by the number of members, minus one). In this case, write performance
is improved. The number of disk drives required to provide availability
can
be unacceptable if arrays are too small.
Notes: - For optimal performance, use arrays with between
6 and 8 member disks.
- When creating RAID arrays with mirroring,
the number of component disks
in each array does not affect redundancy or performance.