RE: [nsp] REG: MPLS VPN route exchange

From: Scott Morris (smorris@mentortech.com)
Date: Fri Oct 06 2000 - 08:26:18 EDT


Regardless of the customers.... You're using MBGP and it's logic to find
only one important thing: The Next Hop!!! So your PE with X & Y makes
decisions on packets coming in. When a packet comes in an interface, that
interface (the serial link from customer X) is associated with a VRF. That
VRF gives the router a "limited" routing table to make its lookups in.

It makes a lookup, picks the next hop (the PE on the other side) and THEN
kicks into the normal MPLS mode... So it would look up in its Label
Information Base what label/exit to take to get to the next-hop address.

The LDP will work like it always does... Either pre-establishing labels, or
as traffic goes, LDP through each hop... However your MPLS is set to work
(works with RSVP too!).

To associate other members, the MBGP extended community is the important
part. In the examples, the RD and extended community all match, which again
is perfectly fine, and simply a design choice on your end. But they don't
HAVE to match. The community string is the important part ACROSS the MBGP
cloud, because that tells everyone what group routes are a part of. The
RD, while technically locally significant only has to map to the correct
community string. IF you make them match, it'll just make your life a bit
easier! :)

The 3600 will fully support MPLS as of 12.1(4)T from what I've heard. Just
from a speed standpoint, I don't believe the 2600 is on the list to be an
LSR at all... Ask your Cisco rep though about that...

Scott
  -----Original Message-----
  From: Vinod Anthony Joseph Cherunni [mailto:vac@dsqworld.com]
  Sent: Friday, October 06, 2000 6:03 AM
  To: smorris@mentortech.com
  Cc: eosborne@cisco.com; cisco-nsp@puck.nether.net
  Subject: RE: [nsp] REG: MPLS VPN route exchange

  Hi,

  Thanks a ton for the guidance, I am sorry for further bothering you with
my queries, Kindly excuse me for it. Just a few queries more.

  My understanding that the IP routing tables are used to build the Label
tables. So assuming my network is built as follows:

  Edge label switch router A - Connecting two customers X & Y. The address
space of X=172.27.x.x, The address space of Y=192.168.129.x.

  Edge label switch router B - Connecting two customers U & V. The address
space of U=192.168.102.x, The address space of V=192.168.104.x.

  Label switch router C connects both the Edge label switch routers acting
as the transit point.

  Now customers X & U form part of VPN -1, & Customers Y&V form part of
VPN-2.

  For example, when Customer X transmits a packet to Customer U who is a
part of the same VPN, ELSR A will label the packet at the ingress, & pass it
on to the LSR C. Now the LSR will need to lookup its Label tabel & forward
the packet to the destination. Now all the core (LSR) will not be aware of
VPN routes, My question is during the label exchange between routers, Will
all the ELSR instruct all the LSR on how the forwarding should happen, since
the ELSR will be aware of the entire IGP routing topology, & will exactly
know how to get to the destination.

  How do we associate members into a VPN, As mentioned in the above
assumption, how will I be able to associate or configure customers X&U to be
part of the same VPN. Because sine RD has only local significance, how will
the MPLS devices understand who is part of which VPN's. Trying to compare
the Frame Relay association of PVC's.

  We are basically an end - end Cisco network, wherein we have a combination
of 2600, 3600, 7200, 7500 routers, & are in the process of getting the IGX
series. Now will the 2600, & 3600 support all MPLS features such as COS,
VPN, & Traffic engineering.

  Thanks a ton & kind regards,

  Vinod.



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:18 EDT