I'm currently reviewing a design that has multiple management groups (for scale), with a root management group containingno agents or devices at all, but simply connecting to each of the other groups using CMG's.
It's my understanding that data does not traverse through CMG's. It simply allows a single unified view of the alert data of each CMG, so I'm confused as to why you would have this root MG, and all it's supporting infrastructure (OpsDB, DW DB etc) when there's no agents at all.
Wouldn't you just have the other MG's (all containing agents) in a CMG configuration, i.e connecting to a single 'local MG' that also has agents.
Cheers,
E