Multihoming in a BGP EVPN VXLAN Fabric Configuration Guide, Cisco IOS XE 26.x.x and Later

PDF

Multihoming in a BGP EVPN VXLAN Fabric Configuration Guide, Cisco IOS XE 26.x.x and Later

Single-active mode

Want to summarize with AI?

Log in

Overview

Describes the redundancy feature that connects hosts or Layer 2 switches to EVPN VXLAN networks through multiple connection paths.

Multihoming in a BGP EVPN VXLAN fabric is a redundancy feature that

  • provides redundant connections between a host or Layer 2 switch and the EVPN VXLAN network

  • connects hosts or Layer 2 switches to the EVPN VXLAN network through either single-homing or multihoming, and

  • supports RFC 7432 and RFC 8365 for VXLAN encapsulation-based EVPN multihoming capabilities on Cisco Catalyst 9000 Series switches.


Single-homing

Single-homing is a network connectivity method that

  • allows you to connect a host or Layer 2 switch to a single VTEP in the EVPN VXLAN network

  • does not support redundancy in the connection between the host or access device and the VTEP, and

  • results in connection loss when the active link breaks down, making single-homed topologies not always reliable and efficient.

Single-homed topology

The following figure shows a single-homed topology:

Figure 1. Single-Homed topology
Topology of a BGP EVPN VXLAN network for single-homing

Multihoming

Multihoming allows you to connect a host or Layer 2 switch to more than one VTEP in a EVPN VXLAN network. This connection provides redundancy and allows network optimization. Redundancy in connection with VTEPs ensures that there is no traffic disruption when a network failure happens. Multihomed topologies are more reliant, secure, and efficient than single-homed topologies.

Multihoming operates in single-active and all-active redundancy modes. In both modes, the connected host or access device connects to the VTEPs through a set of links that together represent an Ethernet segment, identified by a globally unique Ethernet segment ID.

This Ethernet segment ID must also be part of the configuration on the VTEP interface that connects the multihomed host or network device. All traffic forwarded between the VTEPs and the host or Layer 2 switch passes through this Ethernet segment.


ESI-MH in single-active redundancy mode

ESI-MH in single-active redundancy mode is an EVPN configuration that

  • allows two VTEPs to act as an ESI-MH pair with only one VTEP forwarding traffic for a particular Ethernet-segment

  • results in multiple VLANs on the physical link with a single link active for a given VLAN, and

  • enables fabric-enabled VLANs and traditional Layer 2 VLANs to co-exist on the Ethernet-segment enabled trunk-downstream access interface from the VTEPs.

Configuration requirements

The single access link must be a physical Layer 2 interface. Physical interfaces can be part of a Layer 2 port channel for increased bandwidth, but member interfaces in the port channel must be connected to the same VTEP in single-active redundancy mode. From the downstream Layer 2 switch, there should be at least two independent port channel interfaces, with each of these interfaces terminating at a unique upstream VTEP.

Single-active redundancy mode is typically used when all the VLANs on a trunk port connected to an access switch cannot be migrated to an EVPN fabric.


Migration between single-homed and multihomed network topologies

Migration between single-homed and multihomed network topologies is a BGP EVPN VXLAN capability that

  • allows transition from a single-homed topology to a multihomed topology

  • enables removal of redundancy from a multihomed topology to move back to a single-homed topology, and

  • requires corresponding changes to the Ethernet segment configuration to prevent traffic loops and traffic blackholing.

Migration requirements

Note

When you migrate from one topology to another, ensure you make corresponding changes to the Ethernet segment configuration. If you change either of the two without making corresponding changes to the other, it results in traffic loops and traffic blackholing.

For detailed migration procedures: