- NVIDIA CUDA architecture - seems to be relatively inexpensive and has a really good looking development community/support. Lots of decent looking books on Amazon doesn't hurt either..
- BEOWULF cluster - looks like we can build something small and custom (http://www.calvin.edu/~adams/research/microwulf/) and that we'll need to use a parallel library like ( something like MPI). Beowulf seems to be a pretty general term to describe clustering a group of computers from what I understand thus far...
We have software on Linux that we are looking to convert to run on some kind of parallel architecture but honestly we need to do a lot of research before we settle on a good direction to invest our development in. If anyone here knows of any consultants or has any first hand experience in parallel programming or super-computers in general and would either be able to provide some consultation, advice, services, or just some good pointers on the matter - please let me know.