IBGP uses the split horizon rule to avoid loops within an Autonomous System(AS). When an IBGP router receives a route from an IBGP peer, it doesn’t advertise that route to other IBGP peers.

When R2 learns a route from EBGP peer R1, it advertises the route to R3 and R4. However R3 and R4 cannot advertise the same route to R5, hence R5 does not learn routes.

In a network with multiple IBGP peerings, we need to find a solution to ensure full connectivity within the network.
There are three options we can achieve this:
1. Full mesh topology. Every router establishes IBGP peering with every other router. This is not an ideal solution for a Network with many routers, it’s not scalable and hence not commonly used. In a network with N routers, we need to establish N(N-1)/2 IBGP sessions. For example, below networks with 4 routers, we need to establish 6 IBGP sessions 4(4-1)/2.

Imagine you have 50 routers in your network 50(50-1)/2, you need a total of 1225 IBGP sessions. You also need to configure each router one by one. This is why full mesh IBGP is not a scalable solution.

2. Route Reflector: An RR is a designated router that reflects routes learned from one IBGP peer to all other peers. All routers only establish peering with the RR. It’s recommended to have multiple RRs in a large network to avoid a single point of failure.

The broad categories of iBGP routers are Route Reflectors, Route Reflector clients, and Route Reflector non-clients. 

There are a few rules that BGP Route Reflectors obey when it comes to route propagation:

• Routes learned from Route Reflector clients are advertised to other RR clients and RR non-clients.
• Routes received from RR non-clients are advertised to RR clients only.
• Routes learned from EBGP neighbors are reflected to all RR clients and RR non-clients.
• Only the best routes are reflected if there are multiple possible routes.
• The RR cannot change any attributes, including the next-hop address.

3. BGP confederation: Divides the AS into sub-ASes to reduce the number of required IBGP peering. The nested AS also form EBGP peering between themselves.

BGP routers within a sub-AS establish IBGP peering. An RR or full mesh can be implemented within a Sub-AS to ensure full connectivity. BGP routers in different sub-AS establish EBGP peering which means that the AS number is prepended when an update travels between the sub-AS.

In the sample network shown below, we have two Sub-ASes, 65501 and 65502. EBGP is configured between the two Sub-ASes.