RES: [c-nsp] Different Traffic thru BGP links
Zaheer Aziz
zaziz at cisco.com
Tue Nov 23 20:38:40 EST 2004
At 05:07 PM 11/23/2004 -0200, BRA-SAO-Tomaiz,Anderson Goncalves wrote:
>Well..I'm gonna try to explain better. Tell me if I'm not clear again ;)
Now I have a much better picture of your network. I have to work with few
assumptions/questions
1)your customers are not on the same router where Provider A and B are
2)On the box where you have Provider A and B, are there other external
connections?
3)Your demandy Customer is connected to router name C1 and Provider A and B
are connected on
router P
You are absolutely right about using PBR as a solution but expensive one.
Architecturaly there
are two other solutions I can think of
1)Use of MPLS to change forwarding behavior other than best route in your
routing table
2)Encapsulate your traffic in another header (GRE) that you can direct over
desired link
1) may not be possible because if MPLS network ends at your router C, you
can control how traffic
comes from P to C using MPLS but cant control how to choose Link 02 and not
Link 01 to reach customer.
Until you run Inter-AS MPLS TE
http://www.cisco.com/en/US/products/sw/iosswrel/ps1829/products_feature_guide09186a008026468c.html#wp1027129
In this case you would have to make sure that traffic on P goes on Inter-AS
TE Tunnel which traverses Link02.
Caveate:
- All traffic from router P(including Provider A and B) will take
this tunnel to reach Customer Link 02.
- TE Tunnel configuration in your and customers network
2)Here is my simple solution
- Run GRE tunnel from router P to Customer router(actual customer
router and not C1)
- run EBGP on that Tunnel
- Announce Customer prefixes on that EBGP session running on the
tunnel
- On your C1 router create a static router for GRE tunnel end
point to use Link 02
Caveate:
-All traffic from router P(including Provider A and B) will take
this tunnel to reach Customer Link 02
- GRE configuration on Customer router.
I Hope this helps and I dont see any solution other that PBR
Thanks
Zaheer
>I have two internet peerings with providers A and B that I'm exporting
>their routes for all my customers (plus all the other internet routes).
>
>Today one of my customers has one Link with me (let's call link01) that
>I'm exporting all the internet routes to him and I'm receiving his
>prefixes and announcing them to all my peerings (too many than providers A
>and B) and other customers.
>
>This customer wants a new link (let's call link02) and he wants the
>traffic in the link02 will be only traffic between his AS and the AS from
>providers A and B. Remember, that I will keep exporting routes from this
>two providers to other customers. The link01 will have all other internet
>traffic with providers different from A and B.
>
>In customer side is quite simple to solve this using local-preference to
>select which routes will prefer such link (also in link02 I'm announcing
>only routes from A and B to my customer). But in the reverse side, I mean,
>from my backbone to customer way, how do I do this? The prefixes that the
>customer announced to me are the same in link01 and link02 and I cannot
>use local-preference since the rule will be for all the traffic, not only
>from the origineted from providers A and B.
>
>So that's my question..is there a way to solve this without using Policy
>Based Routing(PBR) doing the forward of the packets by the source IP
>address which is not desirable cause both links are 100Mbps and it will
>consume a lot of resources from the equipment.
>
>Any ideas?
>
>Thanks in advanced!
>
>Anderson
>
>
>-----Mensagem original-----
>De: Zaheer Aziz [mailto:zaziz at cisco.com]
>Enviada em: terça-feira, 23 de novembro de 2004 14:29
>Para: Brian Turnbow
>Cc: Zaheer Aziz; BRA-SAO-Tomaiz,Anderson Goncalves;
>cisco-nsp at puck.nether.net
>Assunto: RE: [c-nsp] Different Traffic thru BGP links
>
>
>At 04:43 PM 11/23/2004 +0100, Brian Turnbow wrote:
> >
> >If your using 2 routers twords the customer you can set up traffic shaping
> >tunnels to each router using MPLS, one for the direct peer traffic, one
> >for internet and
> >use local weight to route twords the customer, that way you could still
> >have redundancy
>
>if you use VRFs to separate link A,B, and C then it is difficult to fall
>back to global routing table that is where the redundancy issue comes up.
>Perhaps Anderson should explain his topology a bit more(diagram) so we
>would all be on the same page. Currently we are not.
>
>Thanks
>Zaheer
>
>
> >-----Original Message-----
> >From: cisco-nsp-bounces at puck.nether.net
> >[mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Zaheer Aziz
> >Sent: martedì 23 novembre 2004 16.08
> >To: BRA-SAO-Tomaiz,Anderson Goncalves
> >Cc: cisco-nsp at puck.nether.net
> >Subject: Re: [c-nsp] Different Traffic thru BGP links
> >Importance: High
> >
> >At 04:23 PM 11/22/2004 -0200, BRA-SAO-Tomaiz,Anderson Goncalves wrote:
> > >Hi Guys,
> >
> >
> >This is how I understood your problem,
> >
> >You have ISP connections on Link C and D.
> >
> >You have two providers on say link A and B that must use Link C of your ISP.
> >
> >Link D of your ISP must only be used by your customers and others for
> >general Internet traffic.
> >
> >PBR on A and B was your first thought but due to high amount on traffic on
> >Link A and B, it could be an issue.
> >
> >You could use MPLS_VPN and put A, B and C in a VRF but you must understand
> >that you will loose redundancy, for not using link D in case of failures
> >on C. You could solve this with lots of vrf aware statics as backups but
> >it may not scale.
> >
> >If my understanding of your topology is right then Local_pref that others
> >have suggested will not work because all traffic will follow Link C which
> >is what you dont desire.
> >
> >Thanks
> >Zaheer
> >
> >
> > >I have a situation where an ISP (running BGP) will have two links with
> > >my AS backbone.
> > >In one of them it must allow to pass only peering traffic (from two
> > >other providers that I have direct connection) and thru the other one
> > >normal internet traffic.
> > >To route traffic inbound my AS is simple. Only using local-preference
> > >in the BGP at the customer side, but the problem is how to route
> > >traffic outbound my AS to the customer side, since the prefixes
> > >announced are keeped at the same routing table and are the same on both
> > links.
> > >Use PBR is not desirable, cause there are too many prefixes and to much
> > >traffic thru these connections. I'm also running MPLS VPN in my
> > >backbone, if it has some solution based on this.
> > >
> > >Does anyone has a solution or see it before?
> > >
> > >Thanks for help!
> > >
> > >Anderson
> > >
> > >
> > >_______________________________________________
> > >cisco-nsp mailing list cisco-nsp at puck.nether.net
> > >https://puck.nether.net/mailman/listinfo/cisco-nsp
> > >archive at http://puck.nether.net/pipermail/cisco-nsp/
> >
> >_______________________________________________
> >cisco-nsp mailing list cisco-nsp at puck.nether.net
> >https://puck.nether.net/mailman/listinfo/cisco-nsp
> >archive at http://puck.nether.net/pipermail/cisco-nsp/
>
>
>_______________________________________________
>cisco-nsp mailing list cisco-nsp at puck.nether.net
>https://puck.nether.net/mailman/listinfo/cisco-nsp
>archive at http://puck.nether.net/pipermail/cisco-nsp/
More information about the cisco-nsp
mailing list