BGP policies and packet forwarding

Vikas Aggarwal aggarwal at nisc.jvnc.net
Thu Sep 16 00:54:14 UTC 1993


This is a problem which is perhaps solvable via BGP4 or even otherwise...

I have a network of cisco routers (csc3), running cisco IGRP. I carry
the end-site networks in the IGP. The border gateway to the ENSS runs
igrp as well as BGP.

         a----------+-A----------C-+-----ENSS ----- NSFnet
         b---------/  |          |                   |
                      |          |                   |
         c----------+-B----------D                   |
         d---------/              \------ x -------- y



In the above picture, I denote my backbone routers as capital A B C D
and I am using lowercase 'a b c d' for the end sites.

All the backbone routers run IGRP and carry full end site routing
information.

The problem I am having is in routing for/to 'y'

 a, c, x, y  all are belonging to the same company and 'x' desires
 that traffic from 'a' and 'c' to 'y' goes over the x-y private link.

 However, traffic from the non-company sites (b & d) should flow via 
 the ENSS and then to 'y'.

 How do I tell the routers 'C' and 'D' to send traffic from 'a' and
 'c' via the private link x-y and all other traffic via the ENSS ?
 If a packet arrives at D destined for 'y', then I don't see any
 way of *forwarding* it to 'x' or 'C' based on the *source* address.

 I was reading about 'encapsulation'  in the 'bgp-application' draft,
 but has that been implemented yet ?

 Should I not carry the  'a' 'c'  'x' & 'y' routes in my igrp and just
 use IBGP ? If so, then do I need to create an IBGP peer mesh with
 all routers in my network ? Will that give me the capability to route
 traffic based on the *source* network address ?

Lastly, note that I have drawn a simplistic view of the net... in 
actuality, there are a lot more backbone and end routers involved.

Thanks,

 -vikas                                                    (609) 897-7307
 vikas at jvnc.net	                                 ...rutgers!jvncnet!vikas
-------------------------------------------------------------------------








More information about the NANOG mailing list