[c-nsp] OSPF PoP Design

Nick Kraal nick at arc.net.my
Mon Sep 20 04:07:38 EDT 2004


Dear all,

I am currently working on cleaning up our PoP design and would 
appreciate any help on OSPF design for this.

Currently the PoP topology is like this:

      <---|BB router|--->to other BB PoP routers
          | Area 0  |
	      |
     |Customer aggr. router|
     |      Area 3         |


Both routers have 'redistribute connected metric-type 1 subnets' to make 
the infrastructure networks accessible i.e. /30 customer WAN links, 
router loopback addresses. Also 'redistribute static' on the customer 
aggregation router for NLRI to the customer networks. All these networks 
will be visible in area 0. Bad design I know, especially when the 
customer WAN links flap and CPU cycles taken up on RIB checking, etc. 
But we were young and oh so foolish then.

Clean up.

We are now announcing aggregated PoP prefixes via iBGP. The plan is to 
run NSSA between the BB router and the customer aggr. router. Next 
remove 'redistribute static' so the customer aggr. router should only 
receive a default route as an NSSA area.

What is the best method to then provide NLRI to the customer end as iBGP 
will only route customer network traffic to the BB router only.

1. Customer static route on the BB router. Scalability points zero.
2. Redistribute static on the customer aggr. router with 'summary' or 
'area' statement for prefix aggregation.
3. The real way to do this.

Would appreciate any help, hints, web link, blueprints for point #3 above.

Much appreciated.

-nick/


More information about the cisco-nsp mailing list