The MPLS WG Archive

[Andrew Wu <Andrew.Wu@cosinecom.com>]

Title: RE: Multihop EBGP in 2547bis

      |----------- Multihop EBGP -----------------|
      |                                           |
CE-- PE1 ---(AS1)--- ASBR1<===>ASBR2 ---(AS2)--- PE2 -- CE 

Above is a picture that I'd draw from the text.

In the picture ASBR1/2 would not maintain/redistribute
the VPN-IPv4 routes. Instead the ASBR's only
redistribute the suggested /32 routes to reach PE1 & PE2,
so that PE1 & PE2 will be able to establish their
multi-hop, VPN-IPv4 capable EBGP peering, hence being
able to redistribute their VPN-IPv4 routes directly over
the peering session.

In other words, PE1 & PE2 would eventually establish
an EBGP session over which their VPN-IPv4 routes would be
exchanged, w/out the participation/burdening of ASBR1 & ASBR2.

-andrew

-----Original Message-----
From: Vincent Hamrick [mailto:hamrick@coronanetworks.com]
Sent: Friday, December 15, 2000 9:28 AM
To: mpls@UU.NET
Subject: Multihop EBGP in 2547bis

Please help clear up my confusion with multihop EBGP in 2547bis.

In 2547bis, section 10 option c

      c) Multihop EBGP redistribution of labeled VPN-IPv4 routes between
         source and destination ASes, with EBGP redistribution of
         labeled IPv4 routes from AS to neighboring AS.

         In this procedure, VPN-IPv4 routes are neither maintained nor
         distributed by the ASBRs.  However, an ASBR does use EBGP to
         distribute labeled IPv4 /32 routes to the PE routers within its
         AS.  ASBRs in any transit ASes will also have to use EBGP to
         pass along the labeled /32 routes.  This results in the
         creation of a label switched path from the ingress PE router to
         the egress PE router.  Now PE routers in different ASes can
         establish multi-hop EBGP connections to each other, and can
         exchange VPN-IPv4 routes over those connections.

If the VPN-IPv4 routes are not maintained or distributed by the ASBR, then
how does the VRF assigned label make it to the ingress PE in the other AS?  I
assume that they have to get passed on.

         If the /32 routes for the PE routers are made known to the P
         routers of each AS, everything works normally.  If the /32
         routes for the PE routers are NOT made known to the P routers
         (other than the ASBRs), then this procedure requires a packet's
         ingress PE to put a three label stack on it.  The bottom label
         is assigned by the egress PE, corresponding to the packet's
         destination address in a particular VRF.  The middle label is
         assigned by the ASBR, corresponding to the /32 route to the
         egress PE.  The top label is assigned by the ingress PE's IGP
         Next Hop, corresponding to the /32 route to the ASBR.

Using a downstream unsolicited label distribution, the egress PE would
distribute an IPv4 /32 labeled route with itself as the BGP next hop.  The
ASBR would redistribute this into the next AS (reaching the ingress PE).  If
the ASBR uses BGP, then the P routers have no knowledge of the route and the
third label is needed.  If the ASBR injects into IGP, then the P routers have
knowledge of the IPv4 /32 route.

Now, I try to analyze a packet coming into the ingress PE bound for the
egress PE.  First label put on that packet would be the egress VRF assigned
label.  A route lookup with the BGP next hop of the egress PE obtains the
ASBR assigned label and address.  A route lookup with the ASBR address
obtains the IGP label and interface.  Out it goes.

Are the above statement correct?

Thanks tremendously for any enlightenment,
Vincent Hamrick