From: damien (damien@xxxxxxxxxxx)
Date: Fri Dec 22 2000 - 04:59:38 GMT-3
Hi Kevin,
I will give it a go an try explain it in as few lines as possible, but its
is difficult.
If you think of the following IP Backbone.
PE(Router)------------P(Router)------P(Router)-----------PE(Router)
The P (provider) routers have no customers attached and are Core to the ISP.
The PE (provider edge) Routers are where the VPN customers attach. Each PE
Router must have a host route attached with it, a /32 loopback. The IGP used
in the ISP is essentially used for next hop reachability with the ISP, so it
provides reachability for all the ISP Routers, not VPN Routers.
Tags/ Labels are exchanged between each IGP adjacency for a route and this
is used as the switching mechanism not IP lookup. MP-BGP is used at the PE
Routers (it can be used in the Core also). What BGP does is it assigns
Labels to the VPN Routes and exchanges the routes with the Labels with its
other PE BGP peers. So if you look at from normal perspective. The P routers
have a route to the PE Router via the IGP but not to the routes within the
VPN's. The BGP is not resistributed into the IGP.
So think of a packet entering from the left from a VPN, remember no BGP in
the Core. The packet enters and the PE Router looks in its table and sees
the VPN Route with a Label, it pushes this label into the MPLS label stack
between the Layer 2 and Layer 3 header, it then looks up the next hop and
sees the /32 for the PE Router. It pushes the Label associated with the /32
into the top of the mpls Label stack. The packet now travels to the
destination P router. The P routers examine the top label stack for the /32
label and switch the packet (not bothering to look at the Layer 3 info, even
though they do not have a route to the destination), hence no BGP required
in the Core). This gets switched until it reaches the PE Router, it strips
of the Top lable, then examines the bottom label and sees it has a route for
this label in its VRF table and it routes the packet.
I know is sounds a bit confusing, I would suggest reading RFC2547 and
draft-ietf-mpls-arch-06.txt, they explain it well.
hth
Damien
CCIE
----- Original Message -----
From: "Kevin Gannon" <kevin@gannons.net>
To: "groupstudy" <ccielab@groupstudy.com>
Sent: Friday, December 22, 2000 12:26 AM
Subject: MPLS Removes the need for core BGP tables ??
> I am reading the new Cisco MPLS book and they state that you can use MPLS
> to remove the need for you to have a full BGP mesh within your core
routers
> to achieve transit. Can someone please explain how this actually works
> I have been racking my brains with no luck.
>
> I can understand that you can use a tag pair for the next hop address
> of your upstream peers instead of a tag pair for the actual route. But
> this still doesnt explain how a remote pop will know network 123/8 is
> reachable across the core via POP X.
>
> Regards,
> Kevin
>
>
> --
> ----------------------------------------------------------
> Kevin Gannon
> CCNP,CCDP,MCNS,Cisco Netranger Engineer,HP Openview Consultant
>
> kevin@gannons.net
> ----------------------------------------------------------
>
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:26:09 GMT-3