Advice on "mini" Beowulf cluster

[Patrick Geoffray]

Hi Miska,

Miska Le Louarn wrote:

> Eventually this will be the limiting part in the simulation. We want to
> process the biggest matrix we can afford to. We will use Scalapack (or
> something similar).
> I suspect this might be network intensive, since chuncks of the matrix
> need to be passed to other nodes during the calculation.
> Typically the matrix size will be 10 000 elements (doubles) squared and
> larger.

Matrices multiply of 10000 can easily fit on one machine. Have you 
looked at 4-way SMPs ? 
Scalapack is "dense", you will be fine with Fast ethernet, specially 
for a small number of nodes. Don't use a tiny blocksize and you won't 
produce much traffic (rule of thumb: use the optimal blocksize 
computed by ATLAS and try 1x, 2x, 3x, ...).

If you use a "sparse" package, this is a little different but if 
cost is an important parameter, Fast Ethernet is your best choice.

> I chose Gigabit to be on the safe side. But is that overkill ? 

Yes. For this context, Fast Ethernet is enough. You can still upgrade 
later if you need, but I think you won't. And Fast Ethernet is free 
these days.

Patrick

----------------------------------------------------------
|   Patrick Geoffray, Ph.D.      patrick at myri.com 
|   Myricom, Inc.                http://www.myri.com
|   Cell:  865-389-8852          685 Emory Valley Rd (B)
|   Phone: 865-425-0978          Oak Ridge, TN 37830
----------------------------------------------------------