OSPF design choice: stub or totally stubby?

[Kevin Dorrell]

I am studying the behaviour of the various types of OSPF area. I am happy with the difference between a stub area and a totally stubby area, but to recap .... In a stub area, the ABR does not pass any E1/E2 routes into the area, but gives it a default route instead; however, it still gives it the IA routes. In a totally stubby area, the ABR does not even pass it the IA routes, but relies on the injected default route instead.

My question is this: in a practical design, when would you choose to have a stub rather than a totally stubby? I can see that if there is only one ABR, then there is no point in giving the area all the IA routes; the default route is sufficient, so the appropriate choice is totally-stubby. If there is more than one ABR, then different subnets could be reached more efficiently via different ABRs; the IA route information is useful, so stub would be the appropriate choice. Is it as simple as that: 1 ABR -> configure as stub, more than 1 ABR -> configure as TS (no-summary)? Or are there other criteria that influence the choice?

There is another behaviour, of NSSA, that took me by surprise, but I shall post that question separately.

Kevin Dorrell

Luxembourg

[Harold Ritter · Accepted solution]

This is a difficult question to answer since there is no right or wrong andswer to it. It really depends on what you are trying to achieve. If the goal is to reduce the size of the LSDB and RIB on branch routers with very little memory then totally-stubby might be a better option whether you have one or many ABRs.

I have used TSA in such cases where there were many branches (hundreds of them) connected to two ABRs, which were also WAN aggregation routers. We costed the link to the first ABR lower on half of the remote routers and vice versa for the other half. This provided redundancy and traffic distribution between the two ABRs.

As you said, generally you would use TSA if there is only one ABR. That would not work though if you wanted to deploy MPLS VPN though, as it would break your LSPs.

Bottom line is that it is really hard to say which way to go without knowing what is your final destination.

Hope this helps,

[Kevin Dorrell]

Thanks Harold. Although there may be many answers, at least your response gives me an idea of the sort of things to take into consideration. I like the idea of using the costs to spread the load and provide redundancy, whilst still having the advantage TSA.

Kevin Dorrell

Luxembourg

[Kevin Dorrell]

Just a correction for the record: of course my second paragraph should have read: Is it as simple as that: 1 ABR -> configure as totally-stubby (no-summary), more than 1 ABR -> configure as simple stub?

KJD