Chapter 4
Configuring Multiprotocol Label Switching
Restrictions for BGP Multipath Load Sharing for eBGP and iBGP in an MPLS
VPN
The BGP Multipath Load Sharing for Both eBGP and iBGP in an MPLS VPN feature has the following
restrictions:
•
•
•
Prerequisites for BGP Multipath Load Sharing for eBGP and iBGP in an
MPLS VPN
The BGP Multipath Load Sharing for Both eBGP and iBGP in an MPLS VPN feature has the following
requirements:
•
IGP Convergence Acceleration
From Cisco IOS Release 12.2(33)SB onward, Cisco 10000 series routers support IGP and VPN load
balancing for MPLS VPN scenarios on PRE3 and PRE4 engines. This support allows faster failover of
IGP routes during load balancing. Therefore, for equal cost paths (load-balanced case), convergence is
within an acceptable range.
For unequal cost paths, convergence depends on the number of BGP prefixes; a failover can be more than
30 seconds. From Cisco IOS Release 12.2(33)SB3 onward, Cisco 10000 series routers also support
unequal cost paths.
OL-2226-23
The Cisco 10000 series router supports recursive load sharing, but with the following restriction.
In recursive load sharing, the information required to forward a packet requires at least 2 lookups.
The first lookup determines which provider edge (PE) router is used to reach the final destination.
The second lookup determines how to reach the PE router (from first lookup).
When you configure MPLS VPN, CEF uses recursive load sharing. The first lookup provides the
VPN label, the second lookup provides the IGP label. When PXF forwards a packet, it does only 1
lookup which provides both a VPN and an IGP label; 2 lookups in CEF are combined into 1. The
restriction for recursive load sharing when PXF forwards a packet is as follows.
When there are multiple IGP paths between a Cisco 10000 Series PE router to a provider router (P),
only per-tag load sharing is supported. That is, PXF is programmed with only one of the paths and
this one path is chosen in a round-robin fashion. Because the path is chosen at prefix setup time, it
is not possible to predict which path will be selected for which prefix. The path selected depends on
the order in which the prefixes are configured in the routing table. The bandwidths of the IGP paths
are not considered in the path selection.
When the routing table contains multiple iBGP paths, a route reflector advertises only one of the
paths (one next hop). If a router is behind a route reflector, all routers that are connected to
multihomed sites are not advertised unless separate VRFs with different route distinguishers (RDs)
are configured for each VRF.
Each IP routing table entry for a BGP prefix that has multiple iBGP paths uses additional memory.
We recommend not using this feature on a router with a low amount of available memory and
especially when the router is carrying a full Internet routing table.
MPLS VRFs must be configured before load sharing with both eBGP and iBGP routes can be
configured.
BGP Multipath Load Sharing for eBGP and iBGP in an MPLS VPN
Cisco 10000 Series Router Software Configuration Guide
4-3