FASTORA DAS-315SA Disk Array
115
environments.
A.5 RAID 3
RAID 3 (and RAIDs 4 and 5 also) is based on the
odd parity principle
, which
is best explained through an example. If we assume that there are four
synchronized data streams of bits (“
bi
nary digi
ts
”, which can only be zero or
one), a fifth “parity” stream would be generated, so that the total number of bits
at any given instant is always odd. E.g., if the data bits at an instant in time are
0101, the parity generated would be 1 (total of three bits on); for 0001, p would
be 0 (1 bit on); for 0110, p = 1; for 1000, p = 0; etc. The parity bit is stored with
its corresponding data bits, so that it can be used for either
error detection
or
error correction
.
In
error detection
, if the total number of bits including parity is even, an error
occurred (which bit position is not known without more information). In
error
correction
, in which bit line is in error is known, parity may be generated for all
lines except the one in error, recreating the failed bit stream. Disks are
checked for read failure at the end of each physical block recorded on the
surface of the disk (called a “sector”) when the error checking code (ECC) that
was recorded with the data is compared to the ECC generated on reading back
the data.
In RAID 3, a variable number of data disks (usually four to eight) store blocked
(grouped) data streams (like RAID 0) and a single parity disk records the
generated parity for the data blocks. A data file that is to be recorded is broken
up into blocks and recorded on the data disks, with its corresponding parity
block(s) recorded on the data disks. Because all disks together transfer each
data file, the number of data disks multiplies the data transfer rate, but the
transaction rate is that of a single disk. Both read and write operations involve
