We need an approach here that scales from the smallest...

We need an approach here that scales from the smallest supercomputer up to the very largest,