On Mon, Jun 18, 2001 at 01:17:29PM -0400, Andrew Klaassen wrote:
> > Contrary to popular belief, using both master and slave only
> > gets you a ~5% performance hit.
>
> (?)!
>
> That would be very, very good news. Has anybody else tried
> this? Have you load tested the machines and seen smooth
> performance degradation, no nasty bottlenecks (or whatever it is
> that is supposed to happen when master and slave are both used
> in an array)?
I was curious about this myself and tested it; I found much less
performance degradation than expected with UDMA drives in a RAID-0.
I don't seem to have the results anymore, but off the top of my head,
these two drives typically get 17MB/s and 13MB/s when benchmarked in
isolation; together in a striped volume on different channels, they
max out at 24MB/s, and on the same channel they get 21-23MB/s. Tests
were done on a pair of IBM DeskStar drives (DJNA371350, DTTA371010),
both with on-board VIA MVP3 and a Promise Ultra66 board.
I didn't experience any issues with IDE errors or data corruption.
This was all tested on RedHat 2.2 kernels + patches some time back.
2.4 didn't do so hot at the time, but hopefully it's better now.
So, in summary, it works, and it's not that much slower. YMMV with
different load (my tests were mostly with hdparm and bonnie, and it
wasn't on a RAID-5.)
-ed
|