[WBEL-users] File corruption and new kernel

Tue May 31 11:28:07 CDT 2005

Thanks

I've just found we use the libata driver which is the SCSI framework
version rather than IDE. Are you testing the IDE one?

Our guys in the US say the libata is 50% faster than IDE but slower than
SCSI.

Have you tried this driver. (Don't ask me how you switch to it! :) )

Graham

On Tue, 2005-05-31 at 16:17, Bob Ramstad wrote:
> There may be a slight exaggeration, but I've never run hdparm on /dev/floppy
> 
> With sata_sil enabled on WBEL3 and using /dev/sdX, I get uncached
> reads at nearly 70 Mb/s.  When I go to the built in driver and
> /dev/hdX, it drops to 1.5 Mb/s.  I was wrong, it's more like a factor
> of 45.
> 
> I was stunned at the improvement in performance when I first switched
> over to sata_sil.  Of course, the data corruption is a major drawback,
> but the 0.9 patch seems to have done the trick, given what little
> testing I've managed, and it's pretty clear that the 0.9 patch is
> specific to certain chipsets.  My m/b is nForce2 AMD.
> 
> -- Bob
> 
> On 5/31/05, Graham David Purcocks M.A.(Oxon.) <grahamp at wsieurope.com> wrote:
> > Bob
> > 
> > Is that a *real* speed comparison or are you exagerating a trifle? :)
> > 
> > I'm asking as we use RH and SATA systems and have not specifically
> > noticed poor performance. That said we do have an application which
> > requires a huge disc database (3D model) which runs faster if the data
> > is compressed, which says it is disc bound.
> > 
> > Your statements imply that the built-in driver may have something to do
> > with that.
> > 
> > Not helping your problem but an interesting side issue.
> > 
> > Your comments are appreciated.
> > Graham
> > 
> > On Fri, 2005-05-27 at 17:54, Bob Ramstad wrote:
> > > Righto.  The builtin driver gives 1/20th of the performance of the
> > > sata_sil driver.  Using the builtin driver it's about twice as slow as
> > > a floppy.
> > >
> > > I've run badblocks on every partition using the read write test, and
> > > also run all the diagnostics from the manufacturer, no issues popped
> > > up at all.
> > >
> > > I'm a bit baffled as to what is going on...  but I do see that
> > > sata_sil has been under constant development and I'm sure the version
> > > in the current kernel-source RPM is old.
> > >
> > > -- Bob
> > >
> > > On 5/27/05, Graham David Purcocks M.A.(Oxon.) <grahamp at wsieurope.com> wrote:
> > > > And what happens if you use the crappy builtin driver? Maybe there is a
> > > > reason that is the one being issued.
> > > >
> > > > Don't know but worth asking the question.
> > > >
> > > >
> > > > BTW, We've had problems with a large number of SATA drives being duff,
> > > > so you could have a bad one.
> > > >
> > > > Graham
> > > >
> > > > On Fri, 2005-05-27 at 17:30, Bob Ramstad wrote:
> > > > > Hey there.  Two quickies...
> > > > >
> > > > > 1) I'm running the latest WBEL3 kernel recompiled with sata_sil
> > > > > instead of the crappy builtin driver.  I've been seeing some file
> > > > > corruption, it's odd, but basically when I copy large quantities of
> > > > > files (20+ GB) it is not uncommon for a handful of files to be
> > > > > corrupted.  As an example the other day, I copied 55 GB worth of
> > > > > files, roughly 2200 files in total, and ended up with three that
> > > > > wouldn't pass md5sum --check as compared to the original.
> > > > >
> > > > > Note in this case that the Serial ATA drives were the source, and the
> > > > > destination was a FAT32 filesystem.  Also note that once I recopied
> > > > > those three files, everything was fine.
> > > > >
> > > > > Has anyone else seen anything along these lines?  I would assume it's
> > > > > a Serial ATA specific problem, but not sure.
> > > > >
> > > > > 2) I'm feeling like the best thing to do would be to get a more recent
> > > > > kernel.  Part of me doesn't like doing this as it would fall outside
> > > > > the "yum" model i.e. if there's an update, I won't be automatically
> > > > > notified.  Is there anyone out there aware of a kernel-source yum
> > > > > repository for WBEL3 that is regularly updated for the latest 2.4
> > > > > kernel?
> > > > >
> > > > > -- Bob
> > > > >
> > > > > _______________________________________________
> > > > > Whitebox-users mailing list
> > > > > Whitebox-users at beau.org
> > > > > http://beau.org/mailman/listinfo/whitebox-users
> > > > --
> > > >
> > --
> >
-- 