Load balancing inbound IPSec GRE tunnels

Do you need spoke to spoke communication?
Do you have tiered locations with better connectivity than others?

i.e. site's a, b, c have very high availalbe bandwith T1,DS3
sub-sites d, e, f have partial T's or Frame
the rest have have DSL

An alterntive to hub and spoke might be a mesh which would also provide for convergence in the event a particular network path is not available.

sitea, siteb, sitec will be connected by DMVPN amongst themselves.
Each would have a bunch of spokes and each spoke will have a VPN connection to two or three other spokes of equal bandwidth

using routing protocol i.e. ospf or BGP you can route the inter-spoke and all the way up to any node.

The only thing is that all sites have to have unique LAN segments or you would need to compensate for that using IP transforms/translation.

Actually, I don't want spoke to spoke communication, and, yes, there are variations from site to site regarding bandwidth. To be honest, a peer suggested DMVPN because she'd read something about it but has no direct experience. I will need to do some multicast so that's why I am looking at DMVPN further. The real issue to me is head-end redundancy which I hope to be able to provide with load balancing and IPSec stateful failover.

So . . . load balancing at the head end? You think this can be done?

Thanks for the response Arnold

Steve

Thanks for responding . . .