3com 8807 User Guide
Configuring BGP
381
Configuring BGP Load
Balancing
As BGP is a routing protocol for route selection only, it does not provide a route
calculation method. Therefore, it is not possible to determine whether to enable
load balancing based on a definite metric value. However, the BGP owns a variety
of route selection rules, so it supports conditional load balancing after route
selection, namely, by adding load balancing into the BGP route selection rules.
calculation method. Therefore, it is not possible to determine whether to enable
load balancing based on a definite metric value. However, the BGP owns a variety
of route selection rules, so it supports conditional load balancing after route
selection, namely, by adding load balancing into the BGP route selection rules.
With BGP load balancing enabled, during route selection, the BGP will add rule
between the last two route selection rules "select routes learned from EBGP with
preference" and "select routes advertised by the router with the lowest BGP ID":
if load balancing is enabled, and if there are multiple exterior routes to the same
AS or AS confederation, the BGP will select multiple routes, depending on the
number of routes, for load balancing. The BGP supports load balancing for routes
learned from EBGP/IBGP under the following conditions:
between the last two route selection rules "select routes learned from EBGP with
preference" and "select routes advertised by the router with the lowest BGP ID":
if load balancing is enabled, and if there are multiple exterior routes to the same
AS or AS confederation, the BGP will select multiple routes, depending on the
number of routes, for load balancing. The BGP supports load balancing for routes
learned from EBGP/IBGP under the following conditions:
■
In case of routes learned from EBGP, the BGP performs load balancing for
routes from the same AS and with the same med value only;
routes from the same AS and with the same med value only;
■
In case of routes learned from IBGP, the BGP performs load balancing for routes
with the same med value, local_pref, AS_PATH and origin attributes.
with the same med value, local_pref, AS_PATH and origin attributes.
Figure 91 A schematic diagram of BGP load balancing
As shown in Figure 91, Router D and Router E are IBGP peers of Router C. When
Router A and Router B simultaneously advertise two routes to the same
destination to Router C, if Router C is load balancing enabled (such as balance 2),
while satisfying a certain route selection rule, Router C will add both routes into
the forwarding table if the two routes have the same AS_PATH attribute, so as to
achieve the purpose of BGP route balancing. Router C forwards the routes to
Router D and Router E only once. The AS_PATH attribute will remain unchanged,
but the NEXT_HOP attribute will be changed to the address of Router C, instead of
the address of the original EBGP peer. The BGP will directly transfer the other
transient attributes as attributes of the optimal route.
Router A and Router B simultaneously advertise two routes to the same
destination to Router C, if Router C is load balancing enabled (such as balance 2),
while satisfying a certain route selection rule, Router C will add both routes into
the forwarding table if the two routes have the same AS_PATH attribute, so as to
achieve the purpose of BGP route balancing. Router C forwards the routes to
Router D and Router E only once. The AS_PATH attribute will remain unchanged,
but the NEXT_HOP attribute will be changed to the address of Router C, instead of
the address of the original EBGP peer. The BGP will directly transfer the other
transient attributes as attributes of the optimal route.
The BGP balancing feature also applies to routing between ASs within an AS
confederation.
confederation.
AS200
AS100
Router A
Router C
Router B
Router E
Router D