Beowulf And Digital Signal Processing - Success!

[Huntress, Gary B.]

I've been touting the potential benefits of Beowulf to my colleagues for
the past several months....the lackluster response has basically been
"ok, great, but what can it really *DO*?"

To provide some sort of metric I built a cluster of 10 486's (all
salvaged) with some throwaway NE2000 cards (can you say "sub-optimal"?)
I loaded MPICH and wrote a demonstration program to calculate
spectragrams of acoustic data.  A spectragram is basically a collection
of Fast Fourier Transforms (FFTs).  The original time series is broken
into blocks and distributed among the worker nodes.  Each node performs
the FFTs and the master collects the results.  Very straightforward.  

With 10 nodes (9 worker nodes) I observed a speedup of 8.5 with an 8k
FFT.  I'm very pleased with these results, and I just thought I'd share
a success story.  I strongly suspect that this will lead to more use of
clusters in the DSP area (anyone doing this?  I'd love to hear from
you).

Regards,

Gary Huntress
Code 8313
Naval Undersea Warfare Center
Newport, RI  02841

1-800-669-NUWC  x28990