On Sat, Feb 05, 2000 at 11:39:42PM -0500, Mark Hahn wrote:
> > > dd if=/dev/zero of=/home/1 bs=1024k &
>
> which is the moral equivalent to a fork-bomb: it generates an overwhelming
> stream of requests to the same region of disk.
Ehm, a simple mkisofs does a similar thing...
>
> > It's probably a problem with the elevator sorting in
> > ll_rw_blk. I'm thinking this because the exact same
> > test doesn't give any performance hit at all on my
> > SCSI system...
>
> I think it's merely a fairness issue with allocating request blocks;
> elevator sorting should be fine, since it's c-scan. this isn't an issue
> for scsi because we punt all scheduling to the disk (not always a win...)
>
> regards, mark hahn.
I've installed Andrea's "elevator-starvation-2" patch, and was fine.
Interactive response is much better, apparently without performance losses.
Best wishes,
Gabriele Turchi
-- ---------------------------------------------------------------------------- Gabriele Turchi (Responsabile Tecnico) turchi@tieffesistemi.com Tieffe Sistemi S.r.l. www.tieffesistemi.com V.le Piceno 21 - 20129 Milano - Italia tel/fax +39 02 76115215 ----------------------------------------------------------------------------- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Feb 07 2000 - 21:00:13 EST