I am not sure this is possible, but I'd like to route traffic over an MPLS network using BGP to different locations, but use the same subnet.

We have an east coast and west coast location. Each of these locations has a VPN connection to a common 3rd party location via VPN. The subnet of this 3rd party is We have a number of locations throughout the United States all connected through an MPLS network. I would like those locations to use either the east or west coast VPN.

How could I route some traffic to in one location and some to the other? I read it might be possible using different AS numbers and one would need to coordinate this with the MPLS provider, but really not sure.

Any ideas? If you need more detail please let me know. I am not a network engineer by trade and really just hack my way through this stuff.

Thank you!

  • You seem to be describing anycast: multiple locations with the same addressing, and traffic is sent to the nearest location based on the routing protocol. This is something that is done all the time.
    – Ron Maupin
    Jun 5, 2017 at 22:13
  • Hi Ron, thanks for your input. I am not familiar with anycast, but I took a look at that. Still a little confused on how to do this with BGP and MPLS, but I'll keep digging. Thanks. Jun 6, 2017 at 18:49
  • Basically, a routing protocol can receive a route from several places, but it chooses one as the route to install in its routing table. That is chosen by the metrics used by the routing protocol. The route with the best metrics gets installed in the routing table. Companies do this for things like DNS servers where all the DNS servers have the same address, and a host will then connect to the closest DNS server.
    – Ron Maupin
    Jun 6, 2017 at 18:57
  • Our routers are only receiving routes from one place, BGP from the MPLS provider. Every router is learning the same routes from the same source. So I think I need to work with the provider to manipulate the routes beyond our network. Each router only has one neighbor and that's the path to all the other routers in the MPLS network. It's bifurcating the traffic once it's out on the MPLS network that i'm trying to accomplish. I guess the question is if i'm advertising the same subnet from two location on the MPLS how do I insert a metric into BGP so traffic will route differently over MPLS? Jun 7, 2017 at 14:09
  • It is not your routers that are doing the real routing, it is the provider's routers in the MPLS cloud that must determine the best path. You need to work with the provider to accomplish this. Unfortunately, questions about networks over which you have no direct control are off-topic here. Each provider will have different requirements. It may be that the providers simply uses the number of hops, but that is up to the provider, and you will need to clearly communicate your requirements to it.
    – Ron Maupin
    Jun 7, 2017 at 14:17

5 Answers 5


MP BGP EVPN allow you to extend L2 over any IP network. It encapsulates L2 traffic inside UDP. Although I'm not sure how well it works with large geographic separation. But check it out nevertheless: https://www.juniper.net/documentation/en_US/junos/topics/concept/vxlan-evpn-integration-overview.html

There are also DCI specific technologies such as OTV, which encapsulates L2 inside MPLS inside GRE. This is however a proprietary technology from cisco: http://blog.ine.com/2010/02/15/what-is-overlay-transport-virtualization/

  • Thanks jshen. I'll take a look at these. They do seem a little overly complex for what I'm trying to do, but maybe that's my only option. Jun 6, 2017 at 18:47

Yes it can be done. you can exchange routes using BGP. Now these routes are present in your RIB. For the FIB, it chooses routes based on the administrative distance:- static routes , connected , other routing protocols , etc. if the route exchanged via BGP has the least AD, it will make its way in the forwarding table.

However I doubt it as it is the same subnet, your device will have a route as 'connected' for the subnet. Connected has the least AD. You can still try to use BGP, and the see RI and FIB status.

  • Hi John, thanks for providing this. I tried advertising this route from each location using BGP, but it doesn't appear to be working. When I do a traceroute it just stops out on the MPLS network. It doesn't seem that the AD is working. Is there a way to manually specify this? Is there a way latency can be used to calculate AD? It appears to be the only variable between locations. Thank you. Jun 6, 2017 at 18:53
  • yes administrative distance can be altered. For reference: cisco.com/c/en/us/support/docs/ip/ip-routed-protocols/…
    – john
    Jun 6, 2017 at 19:02
  • Thanks John. The link describes if the router has two different protocols and interfaces to choose from. My situation is that there's only one interface using BGP to the MPLS network. See my comment to Ron above. I think I need to work with the provider to handle traffic differently once it's on the MPLS network. I need a way to insert a metric into BGP that the providers routers will recognize and act on. Is this possible? Thanks again. Jun 7, 2017 at 14:13

After pursuing this with the provider the solution they offered was to switch to a different service they provided, ELAN/VPLS. They stated we could not achieve this on our existing MPLS network. We chose not to pursue further.

Thanks everyone for your input.


You can't advertise prefixes longer than /24 to the internet.

What you can do is advertise the same prefix from many locations. Incoming traffic will go to the closest location and you can then route it within your internal network to where it needs to go.

You may also find a provider willing to do this for you as a service.


Hello you can't advertising same IP from different locations in mpls network. It would cause duplicate IP and drop the traffic in mpls network. Please use other IP range to advertise in mpls network

  • That is perfectly incorrect. You can advertise the same prefix into an MPLS' MP-BGP from more than one edge, as a single customer. This happens usually for customers with multiple PE routers at one site for redundancy. Traffic will not be dropped because of that. Routing logic will act as designed: choosing paths, next hops or egress PE routers according to the given metrics and criteria. However, advertising the same prefix from multiple sites leads to anycast routing, with possibly asymmetric traffic flows, some of which will remain incomplete (perceived as "dropped"). Jul 22, 2019 at 5:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.