EVPN and VRF Support

BGP with EVPN carries virtual network identifier (VNI) and source virtual tunnel end point (VTEP) as part of the MP-BGP standard messaging payload, thereby enabling control plane learning of L2 and L3 reachability information for greater scale of VXLAN networks. As MAC addresses are learned, they are inserted into the EVPN instance table to be propagated by MP-BGP to all other VTEPS in AutoBGP. Learned MACs addresses using BGP are inserted into the local FDB and ARP tables associated with the virtual network. This provides robustness to VXLAN networks with a reduced reliance on flooding when combined with ARP suppression. For a description of EVPN control plane protocol, see RFC 7432. For a description of VXLAN and network virtualization overlay solution using EVPN, see RFC draft-ietf-bess-overlay-08.

EVPN Route Distinguisher and Route Target

A default route distinguisher (RD) per MAC address VRF is auto-assigned per provider edge (PE) as described in Section 7.9 of RFC 7432; Type 1 RD format is the PE loopback address, plus 12-bit Ethernet tag—one MAC VRF per Ethernet tag. Auto-derived route target (RT) is type 0 RD formatted as 2 byte AS number, plus 24-bit network services identifier (NSI) as described in Section 5.2.1 of RFC draft-ietf-bess-overlay-08. The default 2-byte AS number is 64,716. The RD is of local significance only—every VLAN defined on the node has the same number of MAC VRFs associated with it. For example, two AutoBGP nodes 1.1.1.1 and 2.2.2.2 with each having two untagged VLANS 10 and 20. When AutoBGP node 1.1.1.1 advertises EVPN NLRIs from two MAC VRFs using RDs of 1.1.1.1…10 and 1.1.1.1..20, it is possible that VLAN 10 on 1.1.1.1 is associated with VNI 1000 and on 2.2.2.2 VNI, 1000 is associated with VLAN 20. It is the NSI/VNI that brings the VXLAN together on the overlay and ensures that it is properly represented by the RT.

IP VRF Limitations

The following limitations apply to IP VRF: