Information Technology Reference
In-Depth Information
BGP Path Information for 10.2.0.0/16 on R3 (Continued)
Example 5-2
Local
172.16.6.1 (metric 2323456) from 172.16.6.1 (172.16.6.1)
Origin IGP, metric 307200, localpref 100, valid, internal
2323456
Assuming that the default values of parameters such as Weight, LOCAL_PREF, and MED
are not changed, the first point at which there is a difference in the paths is the IGP metric
to reach the prefix's next hop. This is a definite advantage of using iBGP for the network
core. Choosing the best path based on the IGP metric results in the decision's being non-
arbitrary. The path installed causes the traffic to take the shortest path through the network
automatically.
The core routers each choose the path with the lowest IGP metric in a consistent manner.
This does not mean that they all choose the same path. Instead, it means that each router
sends the packet to another router that has chosen the same path, which means that the
destination remains consistent as the packet traverses the network. This behavior results in
deterministic traffic patterns and prevents routing loops from occurring.
If the IGP metrics are the same, the next most common path-selection point is the router ID
of the originating router. In this scenario, no additional information exists about which path
is actually better. It can be argued that choosing the best path based on router ID is arbitrary
because the router ID has no bearing on the path's quality.
The decision is
deterministic
, meaning that the same router is always chosen. However, the
actual path chosen is based on the luck of the draw with respect to router ID, unless the
effect on BGP path selection is taken into account during the address assignment.
NOTE
Other attributes can be used, but care must be taken to ensure that all routers receive iden-
tical information, or there can be potential routing loops. To prevent potential routing loops,
a general rule is to modify attributes only when you originate a route, not between iBGP
peerings.
As soon as the BGP router has selected a path to install into the IP routing table, the router
does a recursive route lookup to determine the IGP next hop. This recursive lookup is done
using the core IGP routing process, because that is the routing process that should contain
the prefix for the iBGP peer where this prefix was originated.
If there are multiple equal-cost paths in the IGP to the BGP next hop, multiple entries
are inserted into the routing table, and traffic is load-shared. Some IGPs, such as EIGRP,
support load sharing over paths with unequal costs. The use of IGP metric for path selection
allows tuning of the IGP link costs to route traffic optimally.
