7

I have one SP right now that serves 100 Sites and we are currently planning to migrate all the sites to a new SP. However, this will be a phased migration, so we will have some sites in the new SP cloud and others in the old one. The SPs will not be running any sort of Inter-AS between them.

What other alternatives do we have to keep communication between sites in the old SP cloud and on the new SP cloud?

EDIT:

The sites are now connected via a single MPLS cloud, there is no IPSec. All the sites connect on a single MPLS cloud. We will be moving some sites to the new MPLS cloud and since there is no connection between the two MPLS clouds (different SPs) and they are not running Inter-AS, the question is what other alternatives do we have to keep sites on, let's say, MPLS-A communicating with sites at MPLS-B.

It is MPLS VPN with BGP routes being exchanged with the PE routers for the Service Provider. So it is basically MPLS L3VPN. Let me know if you need additional info..

3
  • Why would communication not work? NAT? Same LAN?
    – ytti
    Sep 2, 2013 at 12:17
  • 1
    could you add a few more details about how these sites share routing information between them today? Are they connected with IPSec? Are you using MPLS VPN between the sites now? Sep 2, 2013 at 12:19
  • Did any answer help you? if so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you could post and accept your own answer.
    – Ron Maupin
    Jan 3, 2021 at 21:32

3 Answers 3

8

I had this issue. You did not indicate of you have a data center or not. Making the assumption that inter remote site traffic is manageable, in your data centers you allow prefixes to be advertied between the two SP clouds using your data centers as trasits. It is true that remote site to remote side will take the hop through your data centers but there is not much to do. I am also assuming that you will attach your data centers to both providers first, then migrate remote sites the remove your data centers form the old ISP. This will happen by default. If you do not flash cut your remote sites (both providers are connected for a time) use a filter list in BGP to block transit at the remote sites.

If you are using BGP as your routing protocol PE to CE then this will happen almost by default, if not please indicate how you are routing PE to CE.

2
  • Nice one. That's what I thought, we don't have a DC per-se, but we will need to elect a HQ or a main-site and kind multihome that site to both SPs and then all the traffic will transit this site. Is that correct? The problem is that we will need to provision more bandwidth as this site will receive the aggregated traffic of multiple sites. Please let me know if my understanding is correct. Thanks Sep 2, 2013 at 18:18
  • 1
    Yes you are correct. You also want 2 of them (a link can fail). Note that the 'HQ' site will see all the transit between the two MPLS networks, you may be able to schedule the migration where sites that talk to each other a lot move as 'groups' to reduce the amount of transit traffic
    – fredpbaker
    Sep 2, 2013 at 20:36
7

This is a very common situation, which I've seen many times over the years. Regardless of whether you're turning down and replacing legacy Point-to-Point T1 circuits, integrating networks in an acquisition/merger, or just changing providers as in this question, the solution will almost always involve utilizing your existing network equipment/routing configuration for transiting traffic between the different networks. That is the simplest solution and gives you the greatest level of control.

To get on my soap-box for a minute, this is where a clearly-defined, hierarchical network design comes in handy. If you're utilizing a design with Core-Aggregation-Distribution layers (or even a simple Core-Distribution layer setup), you would simply connect each separate MPLS service provider to your Core/Aggregation equipment.

Once the routes to each site (regardless of provider) are in your Core/Aggregation equipment, your network's existing routing protocols can handle the traffic distribution between the two networks. To any router, they will simply be separate subnets available on your network.


Current configuration:

 MPLS Service Provider A
            |
           BGP
            |
       -~-~-~-~-~-
       |100 Sites|
       -~-~-~-~-~-

Desired End configuration:

 MPLS Service Provider B
            |
           BGP
            |
       -~-~-~-~-~-
       |100 Sites|
       -~-~-~-~-~-

And you're seeking the bridge/migration strategy between the two.


EDIT: Clarifying the below solution description per @fredpbaker's comments.

I see two solutions, which both involve connecting your MPLS providers into a core layer at one location (although two would be better for redundancy purposes, and is what is diagrammed below). This would allow your existing network equipment to connect the sites from each provider.

At all other locations/sites, the MPLS provider would be irrelevant. They could be connected to Provider A or Provider B (or both), and traffic would flow between them.

First, you could redistribute BGP into your existing IGP and allow your equipment to route the traffic transiting between each provider:

    Redistribute into OSPF/EIGRP
     /                      \
-~-~-~-~-~-             -~-~-~-~-~-
| Core A  |-\         /-| Core B  |
-~-~-~-~-~-   \     /   -~-~-~-~-~-
    |           \ /           |
    |           / \           |
   BGP       BGP    BGP      BGP
    |        /       \        |
 -~-~-~-~-~-~-       -~-~-~-~-~-~-
 | MPLS SP A |       | MPLS SP B |
 -~-~-~-~-~-~-       -~-~-~-~-~-~- 
             |       |
            BGP     BGP
             |       |
             |       |
             |       |       
         -~-~-~-~-~-~-~-~-~
         | 98 Other Sites |
         |  Connected to  |
         |   Either MPLS  |
         |    Provider    |
         -~-~-~-~-~-~-~-~-~

Secondly, if you don't have an IGP running, or don't want to redistribute into it, you could utilize Inter-VRF routing with VRF lite (as laid out by Stretch over at PacketLife), and take the routes from each provider into a VRF on each Core router, and export them to the appropriate VRF:

Import/Export routes between VRF's as needed
         /                      \
    -~-~-~-~-~-             -~-~-~-~-~-
    | Core A  |VRF-B   VRF-A| Core B  |
    -~-~-~-~-~-   \     /   -~-~-~-~-~-
      VRF-A         \ /         VRF-B
        |           / \           |
       BGP       BGP    BGP      BGP
        |        /       \        |
     -~-~-~-~-~-~-       -~-~-~-~-~-~-
     | MPLS SP A |       | MPLS SP B |
     -~-~-~-~-~-~-       -~-~-~-~-~-~- 
                 |       |
                BGP     BGP
                 |       |
                 |       |
                 |       |       
             -~-~-~-~-~-~-~-~-~
             | 98 Other Sites |
             |  Connected to  |
             |   Either MPLS  |
             |    Provider    |
             -~-~-~-~-~-~-~-~-~
3
  • Your ASCII diagrams are beautiful. And thanks for the plug! Sep 3, 2013 at 14:57
  • maybe I misread the question but it was a migration from SP a to SP b while you present how to do a dual SP MPLS VPN design most sites will not be dual connected. Also not sure what exactly the VRFs do, why not just put OSPF with default infor originate (I have that on 130+ sites that are dual homed to different MPLS providers.
    – fredpbaker
    Sep 4, 2013 at 0:12
  • I can go back and clarify the diagrams a little. My intention for the "98 Other Sites" isn't to illustrate that they're connected to both providers, but that they could be connected to either provider and it wouldn't matter; reachability among all sites would be maintained. As for the VRF's, my first choice would be to redistribute into an IGP from BGP for WAN site-to-Core connectivity. My read of the question/comments seemed to indicate that there was no Core/Data Center, so possibly no existing IGP. I looked for an alternative solution, arriving at VRF-Lite/inter-VRF routing. Sep 4, 2013 at 1:16
1

If you are using Cisco routers, you may want to investigate the LISP protocol

The LISP technology can be easily deployed as an overlay seamlessly interconnecting your two sites across SPs.

There are multiple presentations detailing the most common use-cases but it can easily be extended to your case to provide site extensibility across SPs or even help you in the transition using its host mobility solution.

1
  • 3
    Could you add more details around how LISP answers the OP's question? Sep 3, 2013 at 3:08

Your Answer

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

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