[lug] raid

[D. Stimits]

Peter Hutnick wrote:
> D. Stimits said:
> 
> 
>>There is no technical reason to stop you, it'll work fine, but having
>>slightly different specs can result in far less optimum throughput than
>>with matched disks.
> 
> 
> This is 100% correct.
> 
> 
>>But for raid 5, those in series can be matched, and
>>the one doing checksum would be nice as the 15krpm disk (till you lose a
>> drive).
> 
> 
> This is nonsense.  RAID 5 does not use an independent "checksum" (parity)
> disk.  It alternates, like:

Yes, I had been thinking of placing a pair of raid 0's in parallel, 
using raid 1. But this is raid 1,0 (10), not raid 5, I spoke too soon, 
since the two sets in raid 1 require matching sizes. Looks like he has 
the ability to build the raid, but it will not be at its most efficient.

D. Stimits, stimits AT attbi.com

> 
> Disk1  Disk2  Disk3
> data   data   parity
> parity data   data
> data   parity data
> .. . .
> 
> 
>>You could use the 10k for separate swap, plus other things,
>>like  /tmp/, and gain other benefits. Raid 5 of course is not for
>>performance  reasons anyway, it is for the case of disk failure, so you
>>will probably  be reasonably happy with it.
> 
> 
> This all makes sense, and is good advice if you use RAID-/4/, not 5.  See
> http://www.tldp.org/HOWTO/Software-RAID-HOWTO-2.html.
> 
> -Peter
> 
> 
> _______________________________________________
> Web Page:  http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> Join us on IRC: lug.boulder.co.us port=6667 channel=#colug
>