Running HPC clusters is much easier on the cloud where there’s unlimited capacity and no scheduling problems right? That’s certainly what all the cloud vendors always tell me, but how true is it?