On Tue, 2004-07-13 at 16:36, Alan Dayley wrote:
> George Gambill said:
> > Many years ago, we learned, on an IBM mainframe (slow by today's
> > standards) that we could drastically increase thru put by separating
> > certain files to different disk drives. This reduced head contention.
> >
> > I am wondering if the same distribution of files across a cluster
> > (separate disk drives) might also increase performance. Granted, the
> > drives back then had a 25 ms average seek time and today that number is
> > less than 10.
>
> This is called striping or RAID 0. It is done at a block level now with
> so many blocks of a file going to the first drive, then the next and so
> on. Don't need a cluster of computers to do it, just a RAID controller or
> even software that drives the multiple drives. Linux has a software RAID
> driver already that does various levels of RAID implmentations, including
> RAID 0.
>
> > It's a thought.
>
> It's a good one.
---
RAID 0 is an interesting choice which leaves a system without redundancy
and I would recommend RAID 5 or RAID 10 before I would recommend RAID 0.
This way, you can get the benefits of striping and the benefits of
mirroring too without most of the speed penalty of mirroring and without
the exposure of striping.
Craig
---------------------------------------------------
PLUG-discuss mailing list -
PLUG-discuss@lists.plug.phoenix.az.us
To subscribe, unsubscribe, or to change you mail settings:
http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss
(text/plain)