Instruction manual
171
Cybernetics UM-MV-86-B1-0801
RAID 1 – This is the simplest form of a fault-tolerant array. Based on the concept of mir-
roring, this array consists of multiple sets of data stored on two or more drives. Although
many RAID 1 implementations involve two sets of data (hence the term mirror), three or
more sets can be created if increased reliability is desired.
In a RAID 1 array, storage capacity is determined by the smaller of the two disks in the
array. This provides maximum fault tolerance with complete redundancy but minimum
data storage capacity, since a host system sees only the capacity of a single disk drive.
Writing the same data to the two disks in the array results in poor write speeds, but RAID
1 yields high performance for read-intensive operations.
If a drive failure occurs in a RAID 1 array, subsequent read and write operations are
directed to the surviving drive(s). A replacement drive is then rebuilt using data from the
surviving drive. This rebuilding process has some impact on the array’s I/O performance
because all data must be read and copied from the surviving drive(s) to the replacement
drive.
RAID 1 offers high data availability, because at least two complete sets of data are
stored. Connecting the primary drives and mirrored drives to separate drive controllers
can further enhance fault tolerance by eliminating the controller as a single point of fail-
ure.
RAID 1 has the highest storage cost of any non-hybrid RAID level, because it requires
sufficient drive capacity to store at least two complete sets of data.
RAID 1 arrays are recommended and better suited for small databases or other small-
scale systems that emphasize reliability, such as in accounting or finance.
RAID 10 – This RAID level, originally called RAID 1+0, is implemented as a striped array
(RAID 0) whose segments are RAID 1 arrays. RAID 10 requires a minimum of four disks
installed and is basically several RAID 1 drives linked together with RAID 0. This yields
the speed benefits of RAID 0 with the redundancy benefits of RAID 1. The high I/O rates
are achieved by striping RAID 1 segments. This provides better transfer rate speeds than
RAID 1 alone, especially with write speeds, but not as high as pure RAID 0. This is rec-
ommended for database servers requiring high performance and fault tolerance.
RAID 5 – This widely used RAID type overcomes some of the drawbacks of other parity-
based arrays. In essence, RAID 5 allows for parity information for the array’s data to be
distributed among all drives in the array, a minimum of three, instead of being stored on a
dedicated parity drive.
This distribution parity approach reduces the write bottleneck common to other RAID lev-
els, because concurrent writes do not always require access to parity information on a
dedicated drive. However, overall write performance still suffers because of the overhead
cause by reading, recalculation and updating parity information.
To improve the read performance of an RAID 5 array, the data stripe size can be opti-
mized for the particular application program using the array. Overall RAID 5 array perfor-
mance is equivalent to that of a RAID 3 array except in the case of sequential reads,
which reduce the efficiency of the drives’ read-ahead algorithms because of the distrib-
uted parity information.
As in other parity-based arrays, data recovery in a RAID 5 array is accomplished by com-
puting the XOR of information on the array’s remaining drives. Since parity information is
distributed among all the drives, loss of any drive reduces the availability of both data and