Different DMVPN phases.
When you starting talking about DMVPN you’ll typically hear it being described as a Phase I, II, or III type DMVPN network, so let’s quickly discuss the differences between these three DMVPN phases:
DMVPN Phase I: This phase involves configuring a single mGRE interface on the hub, and all the spokes are still static tunnels so you won’t get any dynamic spoke-to-spoke connectivity. The only advantage of the phase I setup is the fact the hub router’s configuration is much simpler.
DMVPN Phase II: This phase involves everysite being configured with mGRE interface so you get your dynamic spoke-to-spoke connectivity, no more static tunnel destination’s will be configured.
DMVPN Phase III: This phase expands on the scalability of the DMVPN network. This involve summarizing into the DMVPN cloud to provide (Remember EIGRP allows us to summarize out interfaces and BGP allows us to advertise aggregate addresses to neighbors). Along with configuring NHRP redirects and NHRP shortcut switching. NHRP redirects tells the source to find a better path to the destination it is trying to reach. NHRP shortcuts allow DMVPN to learn about other networks behind other DMVPN routers (kind of like ARP for DMVPN). However Phase III will get it’s own post later on.
CCIE or Null! 
I have a question. Hope this is a OK place to ask. We are using running OSPF on shared ipsec NHRP interfaces in a DMVPN topology. The problem we’ve had is that we are unable to set the OSPF cost on the interface because it affects each spoke/peer. Is there a solution for this problem with NHRP? We are actually moving away from our NHRP design due to this limitation. NHRP has worked fairly well and we would love to keep it but it doesn’t seem to scale well considering the OSPF cost limitation. The biggest issue with the cost is that if the remote peer goes down it will try to “load balance” the traffic across multiple sites and bring transfer speeds to a crawl. This also causes issues with our real time streaming application (voIP).
Michael Martinez
September 18, 2014 at 12:40 AM
Of course it is ok!
This is interesting side-effect of a OSPF, have you tried to set the meximum-path to 1 under the OSPF process? By default OSPF performs ECMP over equal cost links you should be able to increase/decrease the number of load balanced links with this parameter.
Is this the only reason you configure OSPF cost of per-spoke tunnel basis?
Stephen J. Occhiogrosso
September 18, 2014 at 7:23 PM
@ Michael: configure ospf network type p2mp non broadcast, this would allow you configure per neighbor cost under ospf to affect the individual neighbor cost and tweak the network flow as you want (you might need to adjust the timers on the spokes to meet the hub or hub to meet the spokes)
senatoredu
February 25, 2015 at 3:56 PM
[…] source for this topic >>> https://ccie-or-null.net […]
Mixed Questions – Network Range
January 14, 2018 at 1:29 PM