|
On this note, does anybody have a line on a good
resource or forum for discussing the RAID4 implementation within the Linux MD
driver? I don't even know if has the capability of running in a
VTL-optimized RAID4 configuration.
For those that don't get it...
Let's say you have 3 disks with disk 3 designated
as the sole parity drive in a VTL-modified
RAID4 configuration.
Let's say that all 3 drives begin as blank and have
a capacity of 4 bits and you want to write the following byte of data to the
array: 10100011
The algorithm requires a read from the parity
drive, a parity calculation based on the new data to be written, writing the
parity back to the parity drive, and a write of the actual data to the first
available data drive.
In this scenario, the following would
happen:
The parity for the capacity to be written to drive
one is read from the parity drive (initially blank):
0000 (parity drive spins up)
The parity is calculated for the first drive write
(let's say using XOR):
1010 XOR 0000 = 1010
Parity is written to the Parity drive:
1010
Data is written to the first drive:
1010 (first drive spins up)
Parity is read for the write to the second
drive:
1010 (still spinning)
Parity is calculated for the second drive
write:
0011 XOR 1010 = 1001
Parity is written to the parity drive:
1001 (still spinning)
Data is written to the second drive:
0011 (drive spins up)
(after a certain level of inactivity, all drives
spin down again)
The End.
So you see, this modified algorithm for RAID4
implementation is optimized for sequential or near-line usage since only the
drive being written to and the parity drive need to be spinning, at the expense
of no longer benefiting from the simultaneous spanned writes available in the
typical RAID4 algorithm.
-=dave
|
_______________________________________________ Openfiler-users mailing list [email protected] https://lists.openfiler.com/mailman/listinfo/openfiler-users
