iBGP

There are cases where we use BGP internally; consider the network diagram below. In this case we treat AS1 and AS3 as spur AS, normally end users who do not require transit.

AS2 is a carrier, providing transit between AS1 and AS3. We will use eBGP between AS1 and AS2 as they are different autonomous systems. The same applies between AS2 and AS3.

However, we need to get the prefixes advertised to Router 2 from Router 1, across to Router 4, so it can advertise them to Router 5. To do this, we use iBGP between Router2 and 4.

Route reflectors

In an iBGP network, all routers must peer with all other routers in a full mesh; this does not scale well. BGP has a concept called a route reflector, which is similar to the notion of a DR/BDR in OSPF.

Every iBGP router peers with the route reflector, which acts as a concentrator for BGP routing information. As a route reflector is a single point of failure, it is normal to provision two at least. A route reflector is a server and any routers peered with it are clients. This would be a simple and effective design for a neutral peering exchange.