[SunRescue] Info on Sun Multipack and disk arrays in general (long)

[Paul Theodoropoulos]

At 05:56 PM 12/3/00, you wrote:
>>>RAID 5: Striping with parity. As I mentioned in RAID one, if a
>>>single drive
>>
>>incomplete - RAID 5 is striping with rotated parity, i.e. the
>>parity blocks are distributed across all the drives in the array,
>
>this is not the way parity works. there is no dedicated parity 
>bit. when parity math is used, any bit can be missing and 
>reconstructed by applying the same mathematical function; thus all 
>bits may or may not be parity and there is not any "rotation" per 
>se.

we're not talking about how parity itself works, we're talking 
about how parity data is distributed. In RAID 5, the parity data 
blocks are distributed across all of the drives in the array, 
rather than being written on a single, dedicated parity data disk.

in the most simplistic graphic:

RAID 3, dedicated parity disk. "XXXX" = Parity data.
'Empty' boxes are user non-parity data.

Disk 1  Disk 2  Disk 3  Disk 4
+----+  +----+  +----+  +----+
|    |  |    |  |    |  |XXXX|
|----|  |----|  |----|  |----|
|    |  |    |  |    |  |XXXX|
|----|  |----|  |----|  |----|
|    |  |    |  |    |  |XXXX|
|----|  |----|  |----|  |----|
|    |  |    |  |    |  |XXXX|
+----+  +----+  +----+  +----+

RAID 5, distributed parity blocks. "XXXX" = Parity data.
'Empty' boxes are non-parity data.

Disk 1  Disk 2  Disk 3  Disk 4
+----+  +----+  +----+  +----+
|XXXX|  |    |  |    |  |    |
|----|  |----|  |----|  |----|
|    |  |XXXX|  |    |  |    |
|----|  |----|  |----|  |----|
|    |  |    |  |XXXX|  |    |
|----|  |----|  |----|  |----|
|    |  |    |  |    |  |XXXX|
+----+  +----+  +----+  +----+

these are lousy examples, but you get the idea.


-----------------------------------
Paul Theodoropoulos   paul at atgi.net
Senior Unix Systems Administrator
Advanced Telcom Group, Inc.
Santa Rosa, California
Work: http://www.atgi.net
Play: http://www.anastrophe.com
Downtime is Not an Option