VXLAN - using an MP-BGP EVN control plane

VXLAN serves as an overlay mechanism for network virtualization. It facilitates the extension of Layer 2 across a common Layer-3 foundation by employing MAC in IP/UDP tunneling encapsulation. This Layer 2 extension within the overlay network aims to bypass constraints imposed by physical server racks and geographical boundaries, offering enhanced flexibility in the positioning of workloads, whether within a single data center or across multiple ones.

The original IETF VXLAN standards, as outlined in the RFC, introduced a multicast-driven flood-and-learn VXLAN model without an integrated control plane. This model depends on data-induced flood-and-learn actions for identifying remote VXLAN tunnel endpoint (VTEP) peers and understanding remote end-hosts. Traffic from the overlay that's broadcasted, unknown unicast, or multicast gets enveloped into multicast VXLAN packets, which are then routed to distant VTEP switches via the foundational multicast forwarding.

However, such a flooding approach may introduce issues with the scalability of the system. Moreover, the necessity to activate multicast features in the foundational network poses difficulties, especially when several entities are hesitant to implement multicast in their data centers or broader network infrastructures.

To overcome these limitations, organizations can adopt the use of Multiprotocol BGP Ethernet Virtual Private Network (MP-BGP EVPN) as VXLAN's control plane. Defined by the IETF, MP-BGP EVPN serves as the standard control plane for VXLAN overlays. It offers a protocol-driven approach for VTEP peer identification and distribution of end-host accessibility data, paving the way for more scalable VXLAN overlays apt for both private and public cloud infrastructures. With the MP-BGP EVPN control plane, traffic flooding within the overlay network is significantly reduced or even eradicated, ensuring efficient traffic routing both laterally (west-east) and vertically (south-north).