8

I am working on a new network deployment, and have run into an interesting question... Should the OSPF area 0 be extended across a WAN's point to point links (thereby creating more border routers), or should one's WAN Aggregation router serve as the ABR for each area? Consider the following:

OSPF Diagram

In Scenario 1, R2, assuming area 1 is a totally stubby area receives only a default route, thereby reducing WAN bandwidth... I think? In Scenario 2, R2 receives many more routes. This seems like a trivial question, but I feel it may have a performance impact at scale.

What is the best practice for extending area 0 across a WAN?

Improve this question
4
  • There is not a "one size fits all" answer here. It depends on the size of your routers, the number of wan links, etc. Can you provide some more details?
    – Ron Trunk
    Jan 27, 2015 at 2:47
  • The WAN aggregation router is an ASR1000 series, and the others are 3945e or similar. There are about 20 links in total, most of the links are 20M metro Ethernet but some of our other sites are T1s (which is where I'm concerned about bandwidth).
    – Justin Jahn
    Jan 27, 2015 at 2:52
  • 1
    Do you run MPLS? Do you run BGP on your WAN routers? Do you run any fast reroute technologies?
    – Daniel Dib
    Jan 27, 2015 at 6:28
  • We don't handle any of the MPLS, our service provider does, so it's a simple layer-2 WAN on our end. We are considering BGP as some of our sites are multi-homed, but there aren't plans to at this time.
    – Justin Jahn
    Jan 27, 2015 at 17:04

3 Answers 3

Reset to default
4

There is no one "best practice," but rather several "good practices" and a few "not-so-good practices."

Generally speaking, if you have multiple areas, you want the hub and spokes in the same area, and summarize between the hub and the rest of the LAN network. You can make the whole hub and spoke area totally stubby.

Another possibility is to break up the spokes into a few (no more than 2-3) areas, with the hub as the ABR. The most important factor that would make this design preferable is the relative instability of your WAN and remote LAN links. If you have a lot of links going up and down, this will affect the amount of flooding over the WAN and therefore the link utilization. But for 50 routes or so, that shouldn't be a problem for a T1.

Here is some more design information.

Improve this answer
1

I prefer scenario 2, extend area 0 into the branch, and then stub area into branch to l3 switch for vlan routing.

Making the branch router ABR.

The reason is that we need high availability for the branches, with a couple of redundant links from different providers, as well as 3G, and dial-up backup.

The bigger branches also have dual routers to cater for a router failure, here we do provision a area 0 link between the routers, using a vlan.

Allowing Area 0 up to the edge of the branch allows for better routing decisions.

We are also able to deploy links between branches that are close together, using the area 0 interface on each branch router, this would not be possible if each branch was in a different area.

Dialup / DMVPN etc also causes problems if the branches routers are only in a area >1 as the central router now needs to have the template/mutipoint-gre in more than one area.

Summary:

I think the WAN should be in area 0, if the wan is not extremely simple/standard with just one or two links to each site.

Improve this answer
1

Scenario 1 would be better in this case since WAN bandwidth will be reduced by a bit and and R2 would be spared of spending its resources if it would be an ABR (scenario 2).

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.