VXLAN Network with MPBGP EVPN Control plane limitation

[davidcheung]

Hi team,

I am reading the Cisco doc "VXLAN Network with MP-BGP EVPN Control Plane"

http://www.cisco.com/c/en/us/products/collateral/switches/nexus-9000-series-switches/guide-c07-734107.html

I know the VLAN ID is locally significant in the context of VXLAN and multiple VXLANs can be mapped to a single VLAN.

But on page 12, from the configuration as the attached:

Looks like each VXLAN has to map to 1 VLAN especially from Distributed IP Anycast Gateway configuration point of view. 

Does this mean that we can only have 4096 VXLAN on each VTEP device?

If not, how we can map multiple VXLANs to 1 VXLAN?  

In addtion, why we need this VXlan_to_vlan mapping? If we temporarily put the integraition with legacy vlan-based network/physical server aside, does a purely VXLAN overlay network not work? 

[Philip D'Ath]

I'm particularly interested in this question.  I wait with baited breath.

[davidcheung]

Anybody from cisco can help?

[Robert Parten · Accepted solution]

I don't work for Cisco, but I work with EVPN VXLAN extensively. For the EVPN VXLAN deployments, yes, you're correct with your presumption for the present time.

Essentially, each VLAN maps to its SVI, which is essentially only tied to a single VRF member (overlay). At present time, only one vn-segment can be configured in each VLAN and, as we all know, an SVI can only be a member of one VRF.

Q-in-VNI offers the "one-to-many" option; however, routing for each subnet/VLAN is obviously done "below" the hardware VTEP.