"Randy Harris" <randy_at_SpamFree.com> wrote in
news:hqMpd.58$nE7.32_at_newssvr17.news.prodigy.com:
>>
>> It's still the standard triangle of cost, benefit and risk
>> with different answers appropriate to different circumstances.
>
> Gentlemen, I believe that you have collectively (along with the BAARF
> site) stated clearly, the case for cost and risk in the triangle that
> Jonathan has identified. I am, however, still having difficulty
> understanding the benefit side, specifically in the area of
> performance between RAID 5 and RAID 10 (or RAID 1 for that matter). I
> do comprehend the advantages to availability, but not performance, and
> hope to be enlightened.
The way RAID-5 works is that when ever a data block is written to a RAID-5
volume a new X-OR block is computed/generated and then this block is
also written to the volume. Under normal operating conditions RAID-5
DOUBLES the number of blocks written to disk. If the disk subsystem
were straining to keep up with activity in RAID-0+1 or RAID-10 setup,
it could not keep up with the additional write activity imposed by RAID-5.
This overhead increases as the number of writes to reads increases to where
it could approach 50% of all disk activity for example if the REDO logfiles
where placed on a RAID-5 volume.
Received on Fri Nov 26 2004 - 15:51:03 CST
