MPLS question..

[mrjdh]

Hello everyone,

I have a GNS3 lab, which was originally created to practice my BGP skills, such as engineering direction of traffic based on certain scenarios. I have since modified the lab to add on 3 routers, connecting to 3 different BGP ASes (e.g. R1 to AS65600, R2 to AS65600, R3 to 65700) so that I can also add in some DMVPN practice.

If possible I would like to modify the lab further to accomodate for an MPLS core setup using MP-BGP, so I can simulate another DMVPN instance for an additional mock company, which use similar IP schemes.

It was then that I had the question...is this even possible to exchange VPNv4 labels using MP-BGP over multiple autonomous systems? I have watched a few hours of labs whereby the MPLS core is within one AS, with PE routers at the edge of the AS, but would like to know whether I can cater for the above setup when there is more than one AS involved? I guess what I'm asking is whether the peers in different autonomous systems can communicate over eBGP sessions but can exchange route information as if they were iBGP peers?

Apologies if this sounds nuts - MPLS and full on VRF weren't part of my ROUTE studies, but would like to get to grips with them. Slightly aware in the past few days that the certificate revamp by Cisco could mean spending time on tech such as MPLS and DMVPN will be wasted! Should have started my Cisco studies earlier in life!

Thanks.

[omz]

Hi 

Technically you can have more than one AS/ISP, different ISPs can route to each other's MPLS networks.

In practice ISPs dont offer this you are a very large customer.

You really don't want to deal with two different ISPs when there is a problem. It will make troubleshooting and getting issues resolved difficult.

[mrjdh]

Thanks Omz.

[Giuseppe Larosa · Accepted solution]

Hello mrjdh,

it is possible to deploy MPLS L3 VPN service over multiple ASes, but this is called inter-AS VPN

inter-AS VPN can be performed using three options:

a)  the ASBR routers use dedicated VRF logical interfaces for each MPLS L3 VPN that needs to be extended and can exchange standard IPv4 in VRF prefixes between them.

This was the first option and actually does not use a MP eBGP session.

This solution is the less scalable as the ASBR needs to be a member or each L3 VPN that needs to be extended to another service provider (requires a VRF importing/exporting route targets of the VRF and an access link that can be a VLan subinterface to the ASBR of the other provider.

b) option B

In option B the ASBR routers between the two SP networks actually use MP eBGP sessions to exchange VPNv4 prefixes.

In this case some tricks are used like next-hop self towards the internal routers.

This is more scalable as the ASBR does not need to be a member of each MPLS L3 VPN that needs to be extended.

c)   use of MP eBGP between Route Reflector servers

So option B is probably what you were thinking of and it is used .

see this article by Daniel Dib

lostintransit.se/2016/03/02/ccde-inter-as-l3-vpns/?doing_wp_cron=1560437724.8246979713439941406250

Hope to help

Giuseppe

[mrjdh]

Wow. Thanks Giuseppe - I'll take a look at Daniel's article.

[mrjdh]

I may come back to ask some more questions..!