The Border Gateway Protocol (BGP) path-selection algorithm prefers external BGP (eBGP) paths over internal BGP (iBGP) paths. With the eiBGP Multipath for Non-VRF Interfaces (IPv4/IPv6) feature, this algorithm is modified to allow multipath load sharing among native IPv4 and IPv6 eBGP and iBGP paths. Prior to the eiBGP Multipath for Non-VRF Interfaces (IPv4/IPv6) feature, this functionality was only available on VPN routing and forwarding (VRF) instances. With this feature, the functionality is extended to non-VRF interfaces. The maximum-paths command allows you to configure BGP to install multiple paths in the Routing Information Base (RIB) for multipath load sharing. The BGP best path algorithm selects a single multipath as the best path and advertises the path to BGP peers. Other multipaths are inserted into both the BGP table and the RIB, and these multipaths are used by Cisco Express Forwarding to perform load balancing, which is performed either on a per-packet basis or on a per-source or per-destination basis.
This feature can be configured on a customer provider edge (PE) device. However, the feature should be configured only on one PE device at the customer site. If this feature is configured on more than one PE device, some parts of the traffic may loop between the PE devices at the customer site. Therefore, it is important to set up the feature appropriately to avoid traffic loops. This feature is enabled by default.