If I may interject with a very hypothetical statement. If you end up using the fog client portion of this solution on more than say a few hundred hosts, you may think about making the server a bit beefier. Here is my setup
For my 1500 hosts that check in with the client:
4 vcpu’s
8GB RAM
250GB storage (we use model specific images)
May be overkill but it works pretty well for me