Showing results for 
Search instead for 
Did you mean: 

OSPF Expert advice

Please take a look at my network. I have the following questions. I would love any suggestions to reconfiguring the ospf domain as this is my weakness.

1. Should I put all connecting interfaces into area 0?

For example I have and in area 20, should this be in area 0?

2. Area 10 and 20 are setup as NSSA and are summarizing routes (redistributed connected subnets) into ospf. Would I get any benefit not using stub or nssa?

3. On Core 1, I am seeing Distro-10 routes advertised as N2 and on Core 2, I am seeing routes from Distro-10 advertised as E2, do you know why? Shouldn't they both advertise as N2?

4. What do I need to do to be able to see 2 equal-cost paths from Distro-10 to Distro-20? If I remove nssa, is it still possible to see 2 equal-path routes over both core1 and core2?

Any help would be greatly appreciated.


I'm going to take a crack at this and if i make a mistake would someone please correct me.....

1. I would have a those areas in area 0. From your layout the only interfaces in area 0 are and You do not have alot of routers be be adding additional areas. The main benifit of areas would be limiting the number of LSA/LSU's.

2. The only benifit from a NSSA is you can have a ASBR while still summerizing routes and limiting the number of LSA you would have in that area. You can still have a ASBR connected to a regular area. The NSSA will inject a default route. You could still summerize routes within area 0. I dont' really see a benifit for using stub or nssa areas with the small amount of routers you have.

3. The default for a NSSA area is N2. An E2 would suggest an inter-area route from the backbone. They both should advertise N2 assuming they are setup properly. I would suggest that you post your configs.

4. I belive that you can us "Default information originate" command on your core 1 and core 2 routers to show they hold 2 routes of equal cost to your Distro 10 and Distro 20 or you could add a cost and have one as preferred.

Could someone else verify what i said...? I working on my BSCI and this is what i have taken away after reading the book one.


Well done :)

Deserves a rating.



Let me share my thought into this along with others...

1.I dont see this as a mandatory as its a small setup.Only difference which comes is in this

case ur ABR is core routers,while if u put ur connected interface also in area 0 then interface connected to core in distr router should also be in area 0 and ur nssa area would be the connected network behind the distribution router .

2.well explianed above :)

3.The reason behind this is when a convertion happens from type 7 to type 5 from nssa area with multiple routers in nssa,that is usually done by a router which is having higher router ID.In ur case that would be Core 1.Now core1 advertise this route into area 0 as type 5 (E2) and core 2 learns this.Now he has learned same route via both E2 and N2.But as the selection process regarding the route type,E2 will be prefered over N2.

Also to add the next hop at core-2 will be still Distr-10,because the forwarding address would be distro-10 router id ,and that is directly reachable..

4.I think the above method is right with nssa.Without nssa i believe if the cost via two path are same then u will see them as equal cost


Thank you all for responding. I will revamp the network and get rid of the areas, I was just planning for future growth.

There are currently 4 different distribution switches connected into the dual core via ospf and their own area.

Each distribution has around 10-12 subnets.

My plan is to setup area 0 for all distribution switches, continue redistributing connected subnets and summarize at the distribution switch up to the core for those routes.

I was reading the cisco data-center design guide and it stated to use nssa and a different area, maybe I was confused.

You weren't confused and actually that's Best Practice. When a company has a data center, it's assumed many sites are collapsing into it. Small networks 'usually' don't need a data center design.



This is a data center not a small network.

There are around 1 thousand servers dual-homed scattered around different distribution switches depending on business unit, etc.