Joseph W. Doherty Thu, 11/15/2007 - 13:00
User Badges:
  • Super Bronze, 10000 points or more

Possible, perhaps, but would it be possible to get both providers to use the same routing protocol?

rsagustin Thu, 11/15/2007 - 14:16
User Badges:

Hi Joseph,

Reference to the diagram. Currently, we are using MPLS from provider-A with OSPF as a routing protocol. But we're bringing up a new MPLS thru a different provider(B). Initially, as a transition steps, it will be load balanced between the two MPLS'on the assumption that it will gonna work. Unfortunately, provider-B can only support BGP. Our main goal is to move away from provider-A for variuos reasons. Eventually, provider-C will be added on the picture minus provider-A. It sounds complex and challenging to me as new to BGP and MPLS environment, thus bringing me to ask help. I would appreciate for any reply I could get. Thanks for your help.

Joseph W. Doherty Thu, 11/15/2007 - 17:00
User Badges:
  • Super Bronze, 10000 points or more

Hopefully since you show your different sites in OSPF areas, do you have a single address block that covers the site? If so, you would advertise that to BGP. You could run BGP and OSPF on each MPLS BGP CE router. Then the CE router would know both topologies.

For the OSPF side, you would need to decide what BGP information to inject. Depending on what you do today, you could inject an OSPF default although that wouldn't lend itself to load balancing.

I'm unsure there will be an easy way to load balance when using a mixture of BGP and OSPF.

What you might also do is, if you don't have too many sites, is build GRE tunnels across the BGP MPLS cloud and run OSPF on the them. This would allow you to cost them as you desire.

rsagustin Tue, 11/20/2007 - 07:49
User Badges:

I don't necessarily need to load balance the two MLPS clouds. I have a new sets of subnets that will be routed thru the MPLS-B. I setup a small lab (see attachment). Again, being new to the BGP and MPLS area, I'm not sure if i configured it at least closer the right way. Somehow, I was able to ping from AS 100 to AS 200 and vice versa if i disconnect either OSPF or BGP links. What do you think with my lab configuration? Your ideas are very helpful. Thanks.

Joseph W. Doherty Tue, 11/20/2007 - 16:04
User Badges:
  • Super Bronze, 10000 points or more

I haven't reviewed your configs with a fine tooth comb, but at first glance they look like you're on the right track.

Some possible things to consider:

On rtE you have BGP sync, probably also want that off.

On rtA you have with BGP maximum-paths 2, don't think you'll need that.

On both rtA and rtE, you have auto-summary. Might not need that.

Why the default-originate on rtD?

On rtD, you might not need the two BGP network statements or neighbor statement other then the remote-AS ones.

On rtA and rtE, probably don't want to redistribute OSPF into BGP but use a network statement that covers the local OSPF area that's behind that router. You also shouldn't need BGP network statements for the BGP connecting subnets.

RtA might need to place interface into OSPF so rtB knows next hop for BGP or you'll need to use next hop self on rtA, I think.

gaurav_thapar79 Thu, 11/22/2007 - 02:01
User Badges:


Can you speak to both upcoming Service providers on OSPF? is it possible for you to migrate onto Single-Area 0 ospf architecture?

If yes, then you can go for OSPF-SHAM link on both MPLS service provider.

By this, you can have dual MPLS -OSPF network and redudant to each other.

Now, load-sharing /balancing, you can play around with ospf-cost.

By this, you can avoid BGP for PE-CE routing protocol.


This Discussion