Cisco Support Community
Showing results for 
Search instead for 
Did you mean: 
Community Member

Link-state protocols design

Hello everyone,

My question might be a bit off topic for this forum, but since most MPLS implementation depend on large scale IGP environment, I thought that it would be a best place to post it.

EIGRP protocol is "flat" in sense that it has no concept of areas/level. You just enable it on a network of any size and summarize IP addresses when necessary.

With OSPF and ISIS it is not clear how large scale real-life deployments are made.

Let say you have a hub-and-spoke topology in your enterprise, so it is advisable to use area 0 on a hub site and regular area on the spoke (stub, nssa are also possible)?

What is one day your two spoke sites connect together? Do you place this new link in the area assigned to one of the spoke sites?

ISIS has the same considerations...

Does anyone have some example topologies from OSPF/ISIS implementations with or without MPLS?

Thanks a lot!



Re: Link-state protocols design

Hello David,

I think using area 0 on the hub is not only desirable but a need.

Also, check this link for some ospf design tips:

Hope this help a bit,

if it does, please rate this post.


Community Member

Re: Link-state protocols design

I think that ISIS is not an issue here, because you can connect the two spoke sites together using L1/L2 link in ISIS and advertiese L2 routes with large metric, so the link between spokes is not used for L2 traffic, while your link to the hub is up. for L1 route optimal roiting can be achieved without carying traffic through the hub

Re: Link-state protocols design


the basic requirement in any MPLS environment is to have LSPs between PE routers in place. Hence from an MPLS perspective the requirement towards IP routing is that the Loopbacks should not be summarized and exist in the routing table as host routes (/32).

You can achieve f.e. by taking all loopback IPs from a separate IP address block and summarize the only rest. This can be done in OSPF multi (normal) area topologies or in ISIS L1/L2 topologies. In the latter you need "redistribute l2 in l1" command, because ISIS by default only inserts from L2 into L1, which would break LSPs.

On the other hand: area size is only limited by memory and CPU load. There are implementations in the real world with nearly 1000 routers in one area (OSPF or ISIS). Those shouldn?t be 836 though ;-)

So multi area can be done, but single area will also scale quite a bit in case carrier class equipment is used.

Hope this helps! Please rate all posts.

Regards, Martin

CreatePlease to create content