beowulf cluster

[error]

> i don't personally have any experience with clustering, but linux
> journal has had a few articles about it recently. and they seem to be
> talking most about an all in one beowulf distro (i think based on red
> hat?) called sycld.
> 
> http://www.scyld.com/page/products/
> 
> and here are some other interesting links i found: 
> 
> http://www.beowulf.org/
> http://www.beowulf.org/howto/howto.html
> http://www.tldp.org/HOWTO/Beowulf-HOWTO.html

Thanks
My main idea at the moment is to set up a debian box as the master PVM
node. At that point I am going to do three things.
The first is set up another box (currently a quad cpu ibm server) to be
a nfs/dhcp/ssh/apt-cache server.
Then I am going to have the option of setting up a netboot floppy that
then mounts the nfs root (different for each kernel image I build).

The third thing is that I am going to do is an automated install (of a
very stripped down debian) with a cd (using the apt-cache server) that
will join the PVM cluster.

I am interested in setting up a boot disk and having all this done via a
NFS mount so that I only have one system to work on (which will matter
when I am dealing with over 60 - 100 boxes).

I really only want the boot floppy to load network card modules and load
a kernel from the nfs server (depending on what it needs). So the node
doesn't even know if it has a drive or a video card. It is just a driver
less node.

So any pointers on that would be great :)

> also, you know this sounds like a pretty interesting topic to speak on
> at some future meeting. ;)

I would be interested in speaking when I actually get the 60 node
cluster running PVM :)

I suggest a field trip to Oakland sometime.
If you want to see the cluster just give me a call first.
Also: wait a while because it is in a state of disarray.

-
Jake