[Beowulf] Raid disk read performance issue in Linux 2.6, anyone seen this?

[Bogdan Costescu]

On Tue, 7 Dec 2004, Craig Tierney wrote:

> I meant White Box distribution which is a rebuild of Red Hat Enterprise
> 3.  It was an early release for Opteron and Itanium.  Since then,
> the Gelato Foundation has taken over the rebuilds and should be
> available.

I don't think that this is "official" WhiteBox Linux. Their web page 
only mentions x86 and x86_86 architectures. They might even don't know 
about it... ;-)
The guy that initially built WhiteBox x64_64 and ia64, Pasi Pirhonen, 
is now their TaoLinux maintainer (along with s390(x)).

> It isn't exact, but every single 2.6 based system I have tried
> has shown this problem.  Every 2.4 based system has not shown this
> problem.

It is indeed strange...

> I primarily used lmdd to test the performance of the filesystem. All
> I care about is big streaming IO.

ext3 and xfs (at least) care about the underlying sector size or RAID
stripe size. Have you paid attention to this when you formatted the
device (if you formatted after moving to the new computers) ? Do you
use something else between the physical device and the file-system,
like lvm (lvm1 in 2.4, lvm2 in 2.6) or software RAID ?

> I did try sgpdd which accesses the device directly and I saw the
> same behavior.

I never heard of this tool. A simple search also didn't found
anything. Care to provide a link ?
Could you run hdparm/zcav/dd/etc. reading directly from real SCSI 
device (no lv*, md*) ? Yes, I know, some of these tools are not very 
precise, but we're talking about almost an order of magnitude...

-- 
Bogdan Costescu

IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu at IWR.Uni-Heidelberg.De