Dec 28 2008

ARF – Automatic Route Filtering

When designing an MPLS network you will have to decide whether to configure a full mesh of MP-iBGP sessions between your PEs as in diagram 1 below or whether to use a hub and spoke topology as in diagram 2 below.

Diagram 1 – BGP full mesh

Diagram 2 – BGP partial mesh

The most obvious benefit of using a hub and spoke topology is that is scales a lot better than a full mesh topology.   As you can see there are a lot less BGP sessions when using partial mesh design.

Automatic Route Filtering

iBGP sessions behave slightly differently with VPNv4 prefix than they do with IPv4 prefixes.  Using diagram 2 as an example, In IPv4 world, R1 would send IPv4 prefixes to R3.  These prefixes would be accepted by R3 and installed into the BGP table.

In VPNv4 world however this is not the case.  In VPNv4 world when prefixes are learned from R1, R3 will reject(filter) the prefixes as R3 is not part of the VPN.  ie R3 does not contain a vrf for those particular VPNv4 prefixes.  This is known as Automatic Route Filtering, and guess what, its turned on by default.  There are two ways to overcome this behavior.  The first is to turn off ARF, the second is to configure R3 as a route-reflector.  When you configure R3 as a route-reflector it turns ARF off by default.

Consider the toplogy below.

We will configure a VPNv4 session  between R1 and R3.  We will create a VPN on R1 called VPNA, which contains the prefix 192.168.1.0/24.  We will redistribute this prefix into MP-BGP which will create a VPNv4 prefix and advertise it to R3.  You will see by default that R3 rejects the route by default.  To accept the route we will have to turn off ARF using the command below.

no bgp default route-target filter

Anyway, take a look at the video below and see ARF filtering the VPNv4 prefix or alternatively download it here and watch it on your iPod.

By zismail • BGP, CCIE SP, MPLS VPN • 3 • Tags: BGP, CCIE SP, MPLS VPN