OSPF interarea route

[bapatsubodh]

Hi,

We have area 10 and area 20 conected to core 0. Both area standard area. ( IA and external routes are available ). A new link of high bandwidth is deployed between two locations of these areas. And we need to push the traffic between area 10 and area 20 through this high bandwidth link instead of the area 0. Is there any way to configure it. Virtual link seems to be one way? Is there any other way. There is a continuous flow of taffic from area 10 to area 20. The main reason to push the traffic throuh this link.

Please share the experience.

Thanks

Subodh

[Giuseppe Larosa]

Hello Subodh,

by protocol design traffic between two OSPF areas has to flow using backbone area

so you need to :

put new link in area 0.0.0.0, to connect the link to the existing ABRs  for (0,10) and (0,20) areas if this is not possible you need to extend the OSPF area 0.0.0.0 topology before using virtual links that I would consider a last resort you can consider the use of logical interfaces in order to have one subif in area X and a new subif in area 0.0.0.0.

This should be easy on lan based links by using VLAN tags

at the end when topology is complete you can play with costs in area 0.0.0.0 topology to have traffic between area 10 and area 20 to go on the new link both sides

Hope to help

Giuseppe

[Peter Paluch]

Hello Giuseppe,

I agree wholeheartedly that the virtual links are not the nicest thing to use, and the proper way would indeed be to have the A10 and A20 ABRs physically join the backbone area. Then again, I believe that a single virtual link would actually solve the OP's problem quite neatly and I personally do not see any significant drawbacks. Can you perhaps elucidate in more detail why you believe that the virtual links should be used only as a last resort, apart from the well-known fact that they do not constitue a best practice?

The IS-IS would make this much simpler, BTW...

Best regards,

Peter

[Giuseppe Larosa]

Hello Peter,

in real world networks, we should make design choices that allow for easy troubleshooting and for tracing traffic.

We shouldn't simply look at elegance or beauty of the solution.

There is a signaling /routing plane and a data forwarding plane

By deploying an appropriate topology you can for example use ACLs to count/trace traffic on the links on the path and so on

You can, depending on device characteristics, use some QoS tools and so on, even if you have only logical links in area 0 and not dedicated links in area 0.

For these reasons, before considering the use of virtual links I would consider the use of point to point GRE tunnels in some cases as a fix that at least has a logical interface countepart on endpoint network devices.

Hope to help

Giuseppe