Least cost routing (BGP) between multiple carriers over MPLS internal network
Bookmark this page:
 Yahoo!
 Google
 Windows Live
 del.icio.us
 digg
 Netscape
|
|
|||||||||||||
|
Posted by Burt Gummer on February 8, 2006, 5:07 pm
Please log in for more thread options each at a different site. We have an ASN and three of our internal routers share BGP tables from the two providers. The remaining three routers are spokes to this core. The "core" is arranged in a triangle of Frame PVCS. So, we have an ISP coming in two two different legs (AT&T on one, Sprint on the other) and the three core routers share the BGP tables. We've being sold a MPLS solution. Our six routers will now be arranged as a full mesh. We're keeping the two ISPs. Currently we use the ISPs as fail-over and Least Cost Routing. Out MPLS vendor says the Label-to-route mapping is actually in the PE and full Internet BGP routing information is too much for them. My understanding of MPLS is limited, so I'm doing a fair amount of reading. I understand, I think, there are mechanisms where the mapping can be moved to the CE, thereby proving us with least cost routing across the full mesh because the mapping burden is in our routers. The PE, then, simply has a table of labels and routes based on those labels we (somehow) attach. Is any of this true? If my understanding s anywhere near correct, please point me to some reference material. My overall impression is my MPLS vendor is trying to screw me by removing functionality from my current network (LCR) and inserting themselves as a "managed" solution, with associative increased cost. | |||||||||||||
|
Posted by Charlie Root on February 10, 2006, 7:03 am
Please log in for more thread options customers + default (possible conditionally set by you based on presence of certain prefix) should be enough, unless you're really into selecting which sites you want to reach via one ISP and which via another. It's quite common for PE routers to have only limited routing table (usually only prefixes within providers own network) as they reach all destination via one or two upstream routers. One of the possible solutions for you (if your providers have such option) is to use multi-hop eBGP towards one of the providers router that carries full-table. > My understanding of MPLS is limited, so I'm doing a fair amount of
Although you indeed could do mapping on your CE, no provider will agree up
> reading. I understand, I think, there are mechanisms where the mapping can > be moved to the CE, thereby proving us with least cost routing across the > full mesh because the mapping burden is in our routers. The PE, then, > simply has a table of labels and routes based on those labels we > (somehow) attach. Is any of this true? > on this due to security reasons (they don't want you to mess up service that they provide). In Frame Relay world it would be comparable if you'd try to get NNI connectivity with your ISP. > If my understanding s anywhere near correct, please point me to some
Managed solution assumes they'd also manage your CE. It's very common among
> reference material. My overall impression is my MPLS vendor is trying to > screw me by removing functionality from my current network (LCR) and > inserting themselves as a "managed" solution, with associative increased > cost. > MPLS VPN providers. Actually it simplifies troubleshooting as there's no ping-pong arguing whos side is at fault. ISP doesn't want to screw you, but they might have not understood what exactly you want to achieve. Try to talk to them and explain why do you want to have full routing and see what they have to offer. Also try first to explain the problem to a manager in your company, so that he/she can escalate the problem with your providers more easier. Regards, iLya | |||||||||||||
> each at a different site. We have an ASN and three of our internal
> routers share BGP tables from the two providers. The remaining three
> routers are spokes to this core.
>
> The "core" is arranged in a triangle of Frame PVCS. So, we have an ISP
> coming in two two different legs (AT&T on one, Sprint on the other) and
> the three core routers share the BGP tables.
>
> We've being sold a MPLS solution.
>
> Our six routers will now be arranged as a full mesh. We're
> keeping the two ISPs. Currently we use the ISPs as fail-over and Least
> Cost Routing. Out MPLS vendor says the Label-to-route mapping is actually
> in the PE and full Internet BGP routing information is too much for them.
>