On Tue, Nov 13, 2007 at 12:18:42PM -0500, Thor Lancelot Simon wrote:
> On Tue, Nov 13, 2007 at 12:22:42PM +0100, Manuel Bouyer wrote:
> > 
> > I suspect it's because of on-board cache issues. As long as there is room
> > in the cache the write is fast, and the controller does the read/modify/write
> > to the disk in background (or, if you're lucky, it ends up with a full stripe
> > and only has to do a write). But once the cache is full, you an I/O needs to
> > wait for some pending writes to complete.
> 
> It seems as if, in that case, the driver could artificially reduce the
> number of openings it presents to the ld layer.  It'd work for SCSI; is
> ld a thick enough layer to make that work?

I don't think reducing the number of openings will help, as the cache is
write-back anyway ...

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--