This goes to show that HP itself doesn't know what they're doing.
On over-provisioning resources:
Your CPU load won't be 100% on all VMs all the time; if it is, you need to scale out anyway. Unused CPU is wasted CPU.
With memory, KSM reduces burden anyway; and again. Unused RAM is wasted RAM.
SSD vs Spinny:
> the host’s fast SSD was pretty much useless, as it had to wait for the 10-times slower HDDs
Huh? Host SSD is meant for the host, not guests, so why is there any 'wait'?
On over-provisioning resources:
Your CPU load won't be 100% on all VMs all the time; if it is, you need to scale out anyway. Unused CPU is wasted CPU. With memory, KSM reduces burden anyway; and again. Unused RAM is wasted RAM.
SSD vs Spinny:
> the host’s fast SSD was pretty much useless, as it had to wait for the 10-times slower HDDs
Huh? Host SSD is meant for the host, not guests, so why is there any 'wait'?
Hold on...
> My examples primarily use Windows
Well... There's your problem...