[email protected]
[Top] [All Lists]

Re: LFAs and OSPF with multiple intra-area AS-external or summary-routes

Subject: Re: LFAs and OSPF with multiple intra-area AS-external or summary-routes
From: Alex Zinin
Date: Wed, 9 Feb 2005 18:14:45 -0800
Yes, I've gone through these too.

The direction I'm going in now is to look at it from the perspective
of intersecting SPTs--each SPT being in it's own area and intersection
points being ABRs. What we need is to make sure that when we move traffic
to an alternate, we don't end up moving it up the tree in another area
and receiving it back to us down that tree.


Wednesday, February 9, 2005, 5:53:44 PM, Alia Atlas wrote:
> Alex,

> One thing to note is that the problem can't occur if D_opt(N, D) < D_opt(E, 
> D).  I.e., if the alternate next-hop is a downstream path with regard to 
> the primary next-hop.  Not that this helps with coverage :-(

> Similarly, if no other ABR in the area is announcing a shorter path than 
> D_opt(E,D), then it couldn't happen.  Or even if the combo of the distance 
> from the other ABR to E plus the path announced is shorter than 
> D_opt(E,D).   If S is an ABR, then S could hear all the other ABR's summary 
> routes and be able to determine this, I think.  Not that it's pretty...

> Alia

> At 08:25 PM 2/9/2005, Alex Zinin wrote:
>> >>Thanks.
>> >>In this scenario, assuming the failure is that of P1 node, Y would lose its
>> >>connectivity in area 1 completely, wouldn't it?
>> > True.  One could easily add a link between Y and another node in area 1 so
>> > that doesn't happen.  There could be a link between S and Y of greater 
>> cost
>> > - so that Y tries to use S as a loop-free link-protecting alternate.
>>I thought about this today. It seems that in the above situation, for a 
>>loop to
>>form, S doesn't even have to be a direct neighbor of Y, just belong to the 
>>rooted at Y's LFA. In fact, it also appears that S doesn't even have to do
>>I need to think more about this, but it seems that the LFA condition for ABRs
>>cannot be simply based on a single area topology. I'll provide more 
>>details when
>>I'm done going through different scenarios.
>>Of course, the last resort is to say that transit areas are not supported yet,
>>but I want to make sure we understand how hard it would be to get them to work

Rtgwg mailing list
[email protected]

<Prev in Thread] Current Thread [Next in Thread>