IBGP is used between PEs to determine routes that are available between VRFs. These routes are linked to a Label Switched Path (LSP) that has been defined separately either as a static path or using LDP or RSVP. The LSP is used to tunnel through the MPLS domain to the destination PE. Under most circumstances, the default route between two PEs is chosen by IBGP between the VRFs with the PEs loopback address as the next hop. When there is a single loopback on the PE, the same LSP tunnel is the only path used between any VRF defined on a PE and VRFs on other specified PEs.
More than one LSP can be configured between PEs however, where each LSP is associated with a different Loopback address on the PE. In this case, any loopback address on a PE can be assigned as the nexthop address for a specific or multiple VRFs. This allows the user to assign some VRFs on a PE to one LSP and other VRFs to a different LSP. Through this method, traffic from different VRFs can be assigned to LSPs that provide different qualities of service. This feature can also be employed to provide for load-balancing across the MPLS domain.
To configure a PE device to use different LSPs, a BGP next hop must be configured for a VRF as the following example illustrates.
(config)# vrf blue (config-vrf-blue)# bgp next-hop loopback 2 (config-vrf-blue)# exit-vrf (config)#
Syntax: [no] bgp next-hop loopback-interface
The loopback-interface variable is the number of the loopback interface that the user is assigning to the VRF as a BGP next hop. The loopback address becomes the defined VRF's nexthop for its VPNv4/VPNv6 routes that are sourced by this device only when:
When these conditions are not met, the default nexthop is used.
For a detailed example of this feature refer to Setting an LSP for Each VRF on a PE.