Information Technology Reference
In-Depth Information
BGP information that is redistributed into the IGP results in a loss of all BGP attributes,
including the AS_PATH. The loss of the AS_PATH attribute defeats the BGP loop-prevention
mechanism. For example, when the prefix is redistributed back into BGP in R4, the same
prefix is sent back to R2, because the AS_PATH contains only 65001. Redistribution also
results in the loss of any policy attributes that have been set for the BGP learned prefixes.
The preferred option is to use internal BGP (iBGP). When R3 advertises the prefixes to R5
via iBGP, R3 does not add its own AS number in the AS_PATH. In fact, Cisco IOS software
does not even check for AS_PATH loops if updates come from an iBGP peer. Without the
additional AS_PATH information, a routing information loop can form within the iBGP
domain.
The loop is avoided if R3 is allowed to advertise the prefix to R5 but R5 is not allowed to
advertise a prefix learned via iBGP to another iBGP peer, such as R4 and R6. However, this
solution requires that all iBGP speakers be fully meshed. For example, R3 is required to
have iBGP sessions with R4, R5, and R6. In an AS that has a large number of iBGP speakers,
a full mesh can present a scalability issue. Solutions to this issue involving route reflection
and confederation are covered in detail in Chapter 7.
The use of iBGP to transport prefix information brings to light another issue. Is an IGP even
needed if BGP can transport all the prefixes?
An IGP is definitely required. In Figure 2-2, R3 is not directly connected to R6. How will
R3 form an iBGP session without some form of routing information to reach R6? The
answer is to have an IGP provide infrastructure reachability inside the autonomous system.
Internal BGP was never designed to exist without an IGP, but in conjunction with an IGP.
An iBGP route is often recursively resolved using an IGP. Table 2-1 shows a few ways in
which iBGP differs from an IGP.
Table 2-1
Comparison of iBGP and IGP
IGP
iBGP
Changes a prefix's next hop at each router to
point to a directly connected address.
Does not change a prefix's next-hop attribute.
Automatically discovers and forms neighbor
relationships.
Requires manual configuration.
Provides information about how to traverse a
given AS or reach a given location.
Provides information about what is available at
a location, without indicating how to reach it.
After you examine the differences between iBGP and IGPs, it is also important to
understand the fundamental differences between iBGP and eBGP (see Table 2-2).
