No, the standby capacity is a physical problem. VM's still need physical compute nodes. If your going to run up 100,000 virtual nodes, asumming each has 2 virtual cpu's on quad processor quad core physical machines, that still means you need need 12,500
physical machines on standby. It could be less if you overcommit the physical machines, but I don't think Azure does and it would kind of suck for high CPU tasks.
I'm saying if you'r selling the promise that you can grab a lot of machines on demand, then you'll have a much higher delta between peek machine use and the low points. And while the physical host machines are sitting unused, they are costing the
hoster money and not making any.
What I'm really questioning, is if we'll ever get to the point that you can at zero notice request 100,000 compute nodes and reilably expect to get 100,000 or if you'll have to book that size request in advance.
Good questions. I'd be disappointed if it doesn't someday become the case. I mean, we are so early in the cloud computing game that one has to assume technological advancements on the order to accomplish what we're talking about here -> "Please process this
on 100,000 CPUs. Thank you."