[c-nsp] IGP & iBGP Configuration Problem in Transit AS

[Nathan]

On Mon, Jun 16, 2008 at 11:20 AM, Vira W <muarwi at gmail.com> wrote:
>
>  My first problem is in IGP configuration inside AS 100. I'm using OSPF. I
> still confused how to make non-BGP router (C,E,B,D) understand how route the
> packet transitting this AS.

You can't. Options:

a) make C E B D speak BGP with all applicable routes

b) change your network topology so that it still works if you remove
all non-BGP-speaking routers (just a connection from A to any of G, F,
or H should do it, but you won't have a lot of redundancy), and make
very sure your IGP doesn't route packets for BGP routers through
non-BGP routers

c) set up tunnels (watch out for MTU problems)

d) implement Layer-2 connections across your backbone using switches
and VLANs and such

e) implement MPLS

Personally I'd go for MPLS, but YMMV.

> Second, I understand that iBGP inside AS 100 needs to be configured in mesh
> topology. Otherwise, use route reflector or confederation. But, if I use
> route reflector, I'm confused because the each route reflector itself must
> be meshed, in the other side, from the physical topology (as in the
> picture), there is none router that is connected mesh. So, which router
> should I choose as RR. Then, if I choose BGP confederation, still inside
> sub-AS the router must be connected in mesh topology, which impossible from
> the topology.

The mesh is logical (having to do with iBGP neighbor configuration on
the routers) not physical (having to do with physical network
connections between routers). In this context a mesh means that every
router must be configured as an iBGP neighbor to all the others (plus
restrictions in option b) above).

-- 
HTH
Nathan