Unanswered Question
Mar 19th, 2009

I need a better information on BGP vs. OSPF

We have a enterprise network, all locations connected through MPLS and DMVPN for internet and BGP was running on both links. Recently one third party guy Mr. Thomas advised my management (who helped at starting this network build) to convert from BGP to OSPF, still I opposed to his advise as....

1. OSPF will consume more bandwidth than BGP

2. Need to create so many areas and the same time update between those areas will become delay

3. OSPF required periodic update where BGP do not required

Here could you give better things about BGP than OSPF and I need your high priority help to prove it at my management.

Thanks in Advance



I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Giuseppe Larosa Fri, 03/20/2009 - 00:29

Hello Naidu,

a) primary links as L3 MPLS VPN

if you are using L3 MPLS VPN as primary links and you use BGP with the PE nodes at each location the current scenario is simpler and doesn't require redistribution between BGP and OSPF.

Moving to OSPF would mean:

need to redistribute BGP into OSPF domain.

becoming external routes the routes on the primary links a second OSPF process is needed on the DMVPN otherwise the backup paths providing O or O IA routes would be preferred to paths from primary links !

This would cause a great increase in complexity.

I would stay with BGP in this case.

b) MPLS links are L2 links or like leased lines

in this case your routers build routing relationships between themselves no SP device involved.

In this case moving to OSPF would be simpler with possible gains in routing convergence.

To be noted that BGP convergence can be improved by using aggressive timers like 2 seconds keepalive 6 seconds timeout.

This can be feasible or not depending on platforms and number of neighbors.

I think the consultant is suggesting OSPF thinking on a routing convergence improvement but there can be an high price in complexity in case a).

case b) here it can be evaluated but BGP convergence can be tuned too.

Hope to help


ilnaiduccna Fri, 03/20/2009 - 04:10

Hi Giuseppe,

Thanks for your respond.

I might not correctly understand your explination, but my openion is.....

Primary links as L3 MPLS

Presently using MPLS as the primary and DMVPN for internet as well VPN tunnels, All PE nodes at each location running with BGP only

Since long BGP were using for our internal routes (one of my location to location) through DMVPN, now it was converted to OSPF and still PE nodes running BGP only

So I would like to know what are the drawbacks about OSPF.



Giuseppe Larosa Fri, 03/20/2009 - 04:18

Hello Naidu,

I would stay with BGP to avoid an increase in complexity of the solution:

if links are L3 MPLS VPN and BGP is used PE-CE you save unnecssary protocol redistributions by using BGP at the central site.

Actually if you move to OSPF at the central site you will see that until you don't divide the OSPF in two OSPF domains you are under the risk that routes from DMVPN are preferred to primary links !!

Note: I don't mean to use different OSPF areas I mean the DMVPN hub routers need to run two OSPF processes: one for DMVPN and one for the central site and to redistribute the first in the second

otherwise if only one OSPF domain is used routes coming from DMVPN can be inter-area O IA and preferred to O E1 or O E2 routes that are the result of redistributing BGP routes into OSPF.

So as you can see the scenario becomes much more complex instead of being simplified.

As I wrote in my first post I would stay with current scenario.

Hope to help


ilnaiduccna Fri, 03/20/2009 - 04:42

Hi Giuseppe Again,

Thanks for your openion with me.

The Scenario as like....

a. Odense is the central site for all DMVPN locations

b. All locations have DMVPN connections and connected to central site Odense

c. Since network defined BGP was running but just before 15 days that was merged to OSPF.

If you menat OSPF domain as I mean Areas then there are 3 areas been created.



lamav Fri, 03/20/2009 - 05:46


Not to jump in unnecessarily, but I am not 100% sure of something but Id like to be so I can appreciate this discussion. It's an interesting one, for sure.

Correct me if Im wrong about any of these points:

1. Your MPLS CE hub router is also your DMVPN hub router. It is in Odense.

2. For the MPLS iinterface, you are running BGPv4 between the hub and spoke sites to create a L3 MPLS VPN.

3. As a backup, you are running DMVPN with OSPF at your hub and spoke sites, of course.

4. The consultant suggests migrating from BGP MPLS to OSPF MPLS.

5. On the core of your network, the IGP is OSPF.

Are all those bullet points correct?



ilnaiduccna Fri, 03/20/2009 - 06:19

Hi Victor,

Nice interaction.

Yes the above mentioned all points are correct.

Preveious the DMVPN also running with BGP but just before 20 days back it was converted to OSPF.

Please go head with investigation and provide me the drawbacks about OSPF as converting from BGP to OSPF.



lamav Fri, 03/20/2009 - 06:42


Giuseppe makes some good points.

Has the consultant explained in detail what the benefits would be of migrating the MPLS environment from BGP to OSPF also?

lamav Sat, 03/21/2009 - 08:20

Oh well, so much for the interesting discussion.

Giuseppe, I think you scared him away. :-) lol

Giuseppe Larosa Sat, 03/21/2009 - 08:33

Hello Victor,

I've reviewed the thread.

if point 5 of your first post (ospf in the core) is correct as confirmed by Naidu I'm starting to think that ,if OSPF can be used as PE-CE protocol, it can be a choice to be evaluated to move to OSPF.

In this case the MPLS Service Provider has to do the right job to make other sites' routes to appear as O IA routes with better metric then DMVPN routes.

If this is true there is no need of the two OSPF domains as I've suggested in my previuos posts.

One possible warning: the MPLS service provider can ask more money for OSPF as PE-CE protocol:

it requires the usage of a dedicated OSPF process on the PE node.

This can be seen as a possible drawback: greater operational expenses.

Hope to help


lamav Sat, 03/21/2009 - 09:23


I have a tendancy to agree with you. Trying to extend the customer's OSPF domain across the provider's MP-BGP backbone requires the proper deployment of OSPF enhancements, like the OSPF Domain community and the ability to pass the LSA type information of the source route, as well as the OSPF cost/MED, to the BGP process.

In that case, the MP-BGP cloud of the SP is going to have to act as a "superbackbone" and logically sit "above" the customer's Area 0 backbone. The PE routers will have to disguise themselves as OSPF ABR/ASBRs and inject inter-area routes on the other side of the MPLS cloud when it redistributes back to OSPF.

In other words, classic redistribution will not suffice because the intra- and inter-area routes will be reinjected back into the customer's OSPF domain on the other side of the SP cloud as external routes, which can really throw off their routing schema. Take note that external routes will remain external routes after redistribution at the distant end.

Given all this complexity, I am wondering why the consultant thought of the migration, which is what I asked Naidu. Im curious to hear his answer.


ilnaiduccna Tue, 03/24/2009 - 02:50

Hi Victor,

Nice interactions being, it's so interesting.

The consultant thought to migrate from BGP to OSPF, as he said to me said that we are using the weight attribute and they may cause to loop.



Giuseppe Larosa Tue, 03/24/2009 - 03:22

Hello Naidu,

>> he said to me said that we are using the weight attribute and they may cause to loop.

This consultant trusts you and your collegues ... :)

Hope to help


lamav Tue, 03/24/2009 - 05:27


Im trying to understand what that means, regarding the loop if you use weights, but I cant.

Anyway, good luck to you, I guess. I hope you seriously investigate what this consultant is saying before you make a drastic move like extending your OSPF domain across the SP's cloud.


ilnaiduccna Fri, 04/03/2009 - 03:16

Hi Giuseppe,

Nice interaction with you it was really help me a lot.

I would like to know more network stuff by you and hope those will help me in my live environment.

NOTE: Is it possible to have you mail address to share the network stuff, if yes it will be so greatfull to me.



Giuseppe Larosa Sun, 04/05/2009 - 11:28

Hello Naidu,

it has been a nice interaction indeed.

My suggestion is to go on on using the forums because everyone here has also a main job, so you can find help from multiple people and share your findings.

The forums are an excellent social network.

if you want to write me you can easily guess my company e-mail:

[email protected]

I can provide you some suggestions/opinions but of course, for time limits, I cannot go into deep details of your projects.

Hope to help



This Discussion