OER PfR limitation on the number of interface

[nicolas.bedard]

Hello,

We have a deployment of 400 store. Each of those have 2 GRE tunnels running over MPLS & 2 GRE Tunnels running over Internet leading to our 2 data-centers. At each Data-Center, we have 1 ASR-1002 connecting both MPLS & Internet MPLS tunnels (800 total per router).

I saw in the documentation that OER & PfR cannot support more than 20 external interface (in our case GRE tunnels) per MC. Does it means that we need to have 20 routers acting as MC to be able to use PfR for our Internet GRE tunnels ?

Is there any more scalable solution for this ? How big company address this issue when they have a lot of interface to run PfR ??

We would only use PfR to make sure that a store get routed toward the best data-center internet link. Currentlly we load-balence manually using OSPF, but this leads to really bad response time when a store happens to connect to data-center #1 even if latency is better at data-center #2.

Thanks

[Steve Lyons]

Hi Nicolas,

Please refer to: http://docwiki.cisco.com/wiki/Performance_Routing_FAQs

specifically: http://docwiki.cisco.com/wiki/Performance_Routing_FAQs#Scale

Q. How many Border Routers are supported, and how many exits in total?
A. A total of 10 Border Routers and 20 exits per Border Router are supported on any one Master Controller for a combined total of up to 200 Interfaces or Exits.

I work in the TAC Routing Protocols Team in RTP and I have seen many customer deployments of PfR. The customers usually deploy 1 MC per site/location.

Let me know if you have any further questions.

Best Regards,

Steve Lyons - Cisco

[nicolas.bedard]

What about having 2 data-center ?

It's easy to configure a branch router to send it's traffic to it's fastest WAN link. You dont have any limitation in regards to the number of interface used.

But when it comes to the returning traffic ? Our store head router has 2 times 400 GRE tunnels. It's way over the 20 linterface limitation.

For what I understand, we would need to add a distribution layer, and have all our MPLS GRE tunnels on one router, ans all our VPN Tunnels on another one. The distribution router would be the one configured with PfR, leading to either MPLS or VPN router. Only 4 interface would be used on the distribution router in this case, but it would not be a cheap solution. We would need 1 new distribution router and 1 other ASR at each datacenter to split our VPN/MPLS into two router instad of one. something like this:

Also, It there a concept of Backup MC ?

What is we have a power outage at a datacenter or if the MC fail for some reason ?

[Steve Lyons · Accepted solution]

Hi Nicolas,
You are correct with your comments regarding support for 2 x 400 interfaces. This far exceeds the current number of supported external interfaces for PfR. You also mentioned you could place a router at the distribution layer and then enable PfR. This would be a viable solution and you might want to consider two routers for redundancy.

PfR does provide non-stateful redundancy/backup where you would enable two PfR MCs and then configure HSRP between them to provide fail over.

If you have one MC and this MC fails then the routing protocols will take over and route based on destination prefixes.

Let me know if you have any further questions.

Best Regards,

Steve Lyons - Cisco