Scyld, Athlon, and Asus A7V?

Keith Underwood keithu at parl.clemson.edu
Sun Jan 7 18:16:19 PST 2001


Glad you got your system up.

Needing step 4 is odd.  It definitely wasn't necessary for us (I had 2
machines up, brought up the third, experimenting with the different BIOS
option, and got better performance).  hdparm works fine for me.

Anyway, glad you got your system up.

					Keith

On Sat, 6 Jan 2001, Jim Phillips wrote:

>
> Thank you!!!  This did it.  Step 4 was actually necessary too - when UDMA
> was set to disabled the kernel would hang, setting it to 4 fixed this.  Of
> course steps 1-3 were also required as far as I can tell.  I also
> commented out the hdparm line in Scyld's node_up script but that was
> before fixing the BIOS so I don't know if it had an effect or not.
>
> In any case, 4 nodes are up, and Scyld is working as promised.  Now if the
> shelves would only arrive we could get the entire cluster online.  ;-)
>
> -Jim
>
>
> On Fri, 5 Jan 2001, Keith Underwood wrote:
>
> > As a matter of fact, I'm slowly putting together 16 1GHz Athlons, same
> > motherboard.  You need to change a couple of things:
> >
> > 1) turn off the virus protection in the BIOS
> > 2) Change the hard disk mode to "User HDD" instead of "Auto"
> > 3) Make sure the addressing mode is LBA
> > 4) Change the UDMA setting to something appropriate (we are using IBM
> > Deskstar drives and the 80 conductor cables, so I chose 4)
> >
> > That last step is just if you want disk performance.  I went from 14 MB/s
> > disk access (hdparm and Bonnie reporting w/ a 1GB file size) to 34 MB/s.
> >
> > 					Keith
> >
> >
> >
> > On Fri, 5 Jan 2001, Jim Phillips wrote:
> >
> > > Hi,
> > >
> > > So, we've just gotten 18 1.1GHz Athlon machines and we're testing out
> > > Scyld.  The one test machine we had worked fine (installed with the Scyld
> > > disk but just ran 1 CPU benchmarks).  Now we're trying to get the slaves
> > > working but the disk partitions keep getting corrupted.  Everything worked
> > > great on an old Pentium 200MHz we had sitting around and we didn't have
> > > any real problems installing the master either.  These are Asus A7V
> > > motherboards with Promise PCI-ATA100 controllers and Western Digital 20GB
> > > drives.  Is there anything in the two kernel monte procedure that could be
> > > causing this problem?  Do we need a newer kernel than 2.2.16?  Is
> > > something wrong with sfdisk running on the slave nodes?  Something creepy
> > > in the BIOS?  Anyone else have experience with these?
> > >
> > > -Jim
> > >
> > >
> > > _______________________________________________
> > > Beowulf mailing list
> > > Beowulf at beowulf.org
> > > http://www.beowulf.org/mailman/listinfo/beowulf
> > >
> >
> > ---------------------------------------------------------------------------
> > Keith Underwood                   Parallel Architecture Research Lab (PARL)
> > keithu at parl.clemson.edu                                  Clemson University
> >
> >
> > _______________________________________________
> > Beowulf mailing list
> > Beowulf at beowulf.org
> > http://www.beowulf.org/mailman/listinfo/beowulf
> >
>
>

---------------------------------------------------------------------------
Keith Underwood                   Parallel Architecture Research Lab (PARL)
keithu at parl.clemson.edu                                  Clemson University





More information about the Beowulf mailing list