EIGRP over top of OSPF

Unanswered Question
Dec 20th, 2007
User Badges:

Our network is divided into two parts (each part is managed separately). Long ago we decided to merge our routing protocols using OSPF. Our side has seen some issues with OSPF because of its inability to load balance the way we want. For example: we have multiple remote networks that have redundant links to out two main hubs. They are not the same distance away as far as the path is concerned. At many times one of the redundant links will be getting hammered while the other link is barely seeing traffic. We would like to start implementing EIGRP to help solve this issue. We would like run EIGRP over OSPF for the sake of the second half of our network. If we really like EIGRP, we will pull OSPF out completely and just use EIGRP and redistribute into OSPF for the second half of our network. The question is this: what kind of behavior could we expect to see while having two routing protocols running at the same time?

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 4 (1 ratings)
royalblues Thu, 12/20/2007 - 08:51
User Badges:
  • Green, 3000 points or more

Well you can always run two routing protocols

The thing to note however is each routing protocol would store its independent database which would require memory and CPU.

Another thing to keep in mind would be to make sure that the AD of the new protocol being introduced be made higher at the beginning so that it does not interfere with the existing routing process. (in ur case EIGRP will take precedence because of lower AD).

After the new protocol learns the network topology and all the routes, migration can be done by reverting back the AD to its normal value.

But as you say that the physical path is not the same distance, it might result in EIGRP behaving the same way. I think it should be quite possible to achieve what is desired by tuning the OSPF metrics.



jamesek Thu, 12/20/2007 - 09:20
User Badges:

I have attached a high level view of a part of our network. The remote sites in the middle with the redundant links to MMC and ncmc-dc-rt01 are the ones that we are having issues with. As you can see from the picture LHC is overusing one link and hardly using the other. What kind of OSPF metrics can be used to help load balance these links? Or what do you recommend?

mohindersingh Thu, 12/20/2007 - 12:01
User Badges:

Hi Jamesek,

OSPF uses the Cost as the metric while comparing two links and different value of cost is associated with different link speeds.

So may be you can set the same link speed on both interfaces bu using bandwidth command and router will consider both links as same capacity and load balance data over both links.

Make sure you are not running QOS on these links. If you are running QOS on those links so QOS will consider teh configured bandwidth as the real bandwidth and divide the traffic among classes according to bandwidth command...

Rate the post if it helps..



jamesek Fri, 12/21/2007 - 08:26
User Badges:

Two things: I am running QoS for the voice traffic over these WAN links. And... Part of the problem is that the shortest route to a destination will usually favor one of the links. The link that is favored is usually the link that is getting hammered. I want a dynamic way to route traffic over the second link when the utilization of the favored link starts to become conjested. Any sugestions?

dblair Mon, 01/07/2008 - 10:38
User Badges:
  • Cisco Employee,

Look at using Performance Routing (PfR) in IOS on the routers terminating the WAN links. PfR

is designed to manage link utilization and application performance across multiple WAN interfaces.

For example, with PfR you can configuring some traffic (voice) to always take the best delay WAN link, while load balancing the other traffic to prevent overloading a WAN link.


chschroe Mon, 01/07/2008 - 17:20
User Badges:

PFR is mostly for exit routing.... to other networks.

PFR is NOT presently optimized for this sort of network - it can only modify packet path with PBR, statics, or BGP, and then we'd have to do a lot of redistribution. Very complicated.


chschroe Mon, 01/07/2008 - 17:17
User Badges:

Can you tell me how EIGRP would solve your issue? Are some of the links different sizes? Where is the traffic going, in general? What sort of traffic is it? Are you not seeing equal cost paths now?

I am trying to figure out how EIGRP would see this different before we really jump into how to run them both. It may not actually fix your problem.

Running EIGRP over OSPF is not really advisable. It can be done, of course, but you will have to change distances on subsets of routes to see some behave differently than others. This is all a VERY complicated process and I really recommend against it.

Lets talk through just the basics before going to the extreme.

shrikar.dange Mon, 01/07/2008 - 22:36
User Badges:
  • Bronze, 100 points or more

considering all above posts, I will like to add one small point to this conversation.EIGRP is capable of load balancing on unequal cost paths where as OSPF load balances on equal cost paths.So when you say that one link is getting hammered while other barely sees traffic check the costs of those links.

you can use (In case of EIGRP) variant command to load balance those unequal cost links.

lets hear more from experts....


shri :)

jamesek Thu, 01/10/2008 - 07:13
User Badges:

The WAN links from our remote sites are equal in cost, but depending on what traffic is being generated their destination probably has a shorter path on only one link. So the preferred path is the one that is having issues while the second path is not being used. The traffic is mostly web traffic or EMR traffic with the occasional large image being moved across the link. Usually it will prefer one of our major hubs over the other because that is the hub that their application is housed at, but for redundancy we felt it was better to have multiple connections.


This Discussion