Users Guide

BGP EVPN for VXLAN
Ethernet Virtual Private Network (EVPN) is a control plane for VXLAN that reduces flooding in the network and resolves
scalability concerns. EVPN uses MP-BGP to exchange information between VTEPs. EVPN was introduced in RFC 7432 and is
based on BGP MPLS-based VPNs. RFC 8365 describes VXLAN-based EVPN.
The MP-BGP EVPN control plane provides protocol-based remote VTEP discovery, and MAC and ARP learning. This
configuration reduces flooding related to L2 unknown unicast traffic. The distribution of host MAC and IP reachability
information supports virtual machine (VM) mobility and scalable VXLAN overlay network designs.
The BGP EVPN protocol groups MAC addresses and ARP/neighbor addresses under EVPN instances (EVIs) to exchange them
between VTEPs. In OS10, each EVI is associated with a VXLAN VNI in 1:1 mapping.
Benefits of a BGP EVPN-based VXLAN
● Eliminates the flood-and-learn method of VTEP discovery by enabling control-plane learning of end-host L2 and L3
reachability information.
● Minimizes network flooding of unknown unicast and broadcast traffic through EVPN-based MAC and IP route
advertisements on local VTEPs.
● Provides support for host mobility.
NOTE: This feature is not supported on the N3248TE-ON platform.
BGP EVPN compared to static VXLAN
OS10 supports two types of VXLAN NVO overlay networks:
● Static VXLAN
● BGP EVPN
Configure and operate static VXLANs and BGP EVPNs for VXLAN in the same way:
● Manually configure the overlay and underlay networks.
● Manually configure each virtual network and VNI.
● Manually configure access port membership in a virtual network.
● Existing routing protocols provision and learn underlay reachability to VTEP peers.
However, static VXLANs and BGP EVPNs for VXLAN differ as described:
Table 97. Differences between Static VXLAN and VXLAN BGP EVPN
Static VXLAN VXLAN BGP EVPN
To start sending and receiving virtual-network traffic to and
from a remote VTEP, manually configure the VTEP as a
member of the virtual network.
No manual configuration is required. Each remote VTEP is
automatically learned as a member of a virtual network from
the EVPN routes received from the remote VTEP. After a
remote VTEP address is learned, VXLAN traffic is sent to, and
received from, the VTEP.
Data packets learn remote hosts after decapsulation of the
VXLAN header in the data plane.
Remote host MAC addresses are learned in the control plane
using BGP EVPN Type 2 routes and MAC/IP advertisements.
VXLAN BGP EVPN operation
The EVPN address family allows VXLAN to carry EVPN routes in External Border Gateway Protocol (eBGP) and Internal Border
Gateway Protocol (iBGP) sessions. In a data center network, use eBGP or iBGP for route exchange in both the IP underlay
network and EVPN.
The following sample BGP EVPN topology shows a leaf-spine data center network where eBGP exchanges IP routes in the IP
underlay network, and exchanges EVPN routes in the VXLAN overlay network. All spine nodes are in one autonomous system—
AS 65535. All leaf nodes are in another autonomous system—AS 65000.
To advertise underlay IP routes, eBGP peer sessions establish between the leaf and spine nodes using an interface IP address.
To advertise EVPN routes, eBGP peer sessions between the leaf and spine nodes use a Loopback IP address.
1156
VXLAN