Mailing List Archive: [nsp] Odd things with BGP and VRF

[nsp] Odd things with BGP and VRF

Jul 31, 2002, 12:49 PM

Post #1 of 3 (1921 views)

Wondering if anyone can tell if this is even possible...

I would like to have routers peer with multiple ISP's, getting a full
routing table from each of them, and iBGP peering as normal.

Internally, I need another system to talk BGP to those routers, and not
just get the normal iBGP route updates that consist of the best path
that router would use, but all the paths it heard from all peers. This
internal box is going to be doing all sorts of analysis of paths and
performance, and it needs the full views as presented by each provider.

I could try and get each provider to set up eBGP multihop to this box,
as well as our regular router peerings, but don't much fancy my chances
of achieving that.
So I was wondering if I can do want I want with VRF's: something like
have each external peer be in a different VRF, so that the tables are
kept separate, but have the internal machine peer with each VRF-bgp table.

Is that possible? Can you have BGP processes running per VRF?

And can one system possibly peer with more than one of them?

TIA

Re: [nsp] Odd things with BGP and VRF [ In reply to ]

will at charnocks

Aug 1, 2002, 1:48 PM

Post #2 of 3 (1873 views)

Permalink

----- Original Message -----
From: "Steve Francis" <sfrancis@expertcity.com>
To: <cisco-nsp@puck.nether.net>
Sent: Wednesday, July 31, 2002 2:49 PM
Subject: [nsp] Odd things with BGP and VRF

> Internally, I need another system to talk BGP to those routers, and not
> just get the normal iBGP route updates that consist of the best path
> that router would use, but all the paths it heard from all peers. This
> internal box is going to be doing all sorts of analysis of paths and
> performance, and it needs the full views as presented by each provider.

I don't think any vendors support this, but there is an Internet draft that
would probably accomplish what you're asking for:

http://search.ietf.org/internet-drafts/draft-walton-bgp-add-paths-00.txt

--
William R. Charnock
Director of Core Technology
Allegiance Telecom, Inc

RE: [nsp] Odd things with BGP and VRF [ In reply to ]

Rene.Koning at versatel

Aug 2, 2002, 12:18 AM

Post #3 of 3 (1890 views)

Permalink

Steve,

You can use VRF's to isolate the ISP BGP peering routes.
Terminate each peer in a different VRF (be sure to get some
extra memory :). Although there is only one BGP process
per router the routes are segregated by Routing Distinguishers.

René Koning

> -----Original Message-----
> From: Steve Francis [SMTP:sfrancis@expertcity.com]
> Sent: Wednesday, July 31, 2002 9:49 PM
> To: cisco-nsp@puck.nether.net
> Subject: [nsp] Odd things with BGP and VRF
>
> Wondering if anyone can tell if this is even possible...
>
> I would like to have routers peer with multiple ISP's, getting a full
> routing table from each of them, and iBGP peering as normal.
>
> Internally, I need another system to talk BGP to those routers, and not
> just get the normal iBGP route updates that consist of the best path
> that router would use, but all the paths it heard from all peers. This
> internal box is going to be doing all sorts of analysis of paths and
> performance, and it needs the full views as presented by each provider.
>
> I could try and get each provider to set up eBGP multihop to this box,
> as well as our regular router peerings, but don't much fancy my chances
> of achieving that.
> So I was wondering if I can do want I want with VRF's: something like
> have each external peer be in a different VRF, so that the tables are
> kept separate, but have the internal machine peer with each VRF-bgp table.
>
> Is that possible? Can you have BGP processes running per VRF?
>
> And can one system possibly peer with more than one of them?
>
> TIA
>
> _______________________________________________
> cisco-nsp mailing list real_name)s@puck.nether.net
> http://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/