Swivel bills as “YouTube for data.” You upload your data and use Swivel’s servers and algorithms to analyze that data and compare it to any other similar data uploaded by others.
The interesting thing for us is that this is unlike YouTube in a key way: these data have to be sliced, diced, processed, and other massaged before presentation. This is implicit HPC as a service. If the company is successful, they are going to have very large (Google-scale) servers on their hands.
It would be interesting to know whether they can do this in a loosely-coupled kind of way, or whether they’ll need HPC big iron to get it done. My bet is that the data sets will likely be numerous and small, making scads of cheap boxes and commodity interconnects a reasonable solution.