Stub AS

Stub AS A stub AS is an AS that does not allow information to transit through it to another
AS. If you refer back to Figure 8.10, notice that both AS 100 and AS 300 are stub ASs. AS 100
and AS 300 are single-homed autonomous systems. A single-homed AS is an autonomous system
that has only one entry and exit point. All single-homed autonomous systems are stub ASs.
BGP synchronization requires that BGP be synchronized with the IGP before any transit
information can be advertised. In other words, the eBGP speaker will wait to hear an advertisement
of the route it learned via iBGP from the IGP running in the AS, before advertising the
route to an eBGP neighbor. To better understand this concept, refer to Figure 8.11.
In this example, R1 and R2 are eBGP peers, R2 and R4 are iBGP peers, R4 and R5 are eBGP
peers, and R3 is not running BGP. What will happen when R1 sends a packet destined for R5 is
that the packet is received by R2, which in turn forwards the packet to R3, because R3 is the IGP
next hop for R2 to reach R4, the BGP next hop. Because R3 is not running BGP and the BGP
routes have not been redistributed into the IGP, R3 doesn’t know how to get to R5. So R3 will
drop the packet. There are a couple of ways to overcome this:
 You could redistribute the BGP routes into the IGP. This would seem like the most logical
thing to do. However, redistributing BGP routes into an IGP is not a good idea. In 1994, there
were under 20,000 routes in the Internet’s BGP routing tables. At the start of 2005, there were
over 180,000 routes in the default-free Internet, so an IGP would have a meltdown if required
to carry that many routes. Never mind the continued growth that is expected.
AS100
AS 200
AS 300
258 Chapter 8  Border Gateway Protocol
 You could run iBGP on R3, which would allow you to disable BGP synchronization. By
doing this, R3 will know that to reach R5 it must forward the packet on to R4, and the route
to R5 will not need to be synchronized with the IGP. Being an iBGP speaker in a full-mesh
configuration, R3 would be informed by R4 that R4 is the BGP next hop to R5 at the same
time that R2 would learn that fact. Therefore, there is no need to wait for the IGP to learn this.
In general, there are two conditions that would enable synchronization to be turned off, if
either or both are true. The first is that the AS is a stub and carries no transit traffic. The second
is if all routers in the AS speak iBGP and are configured in a TCP full mesh with one another.
Because full-mesh iBGP is highly recommended and often configured, most implementations
of BGP you see in the real world will have BGP synchronization turned off. By default, BGP synchronization
is on. In order to turn off BGP synchronization, you need to enter the following
command in router configuration mode: