[email protected]
[Top] [All Lists]

Re: Ordered FIB approach & broadcast links

Subject: Re: Ordered FIB approach & broadcast links
From: Alia Atlas
Date: Wed, 13 Apr 2005 10:24:10 -0400
Mike,

I agree that recognizing the pseudo-node has changed its name would remove this additional concern for handling "good news" and "bad news" simultaneously.
Alia

At 06:19 AM 4/13/2005, mike shand wrote:
Alia,
Thinking about this just a tiny bit more, when we have a failure of a DR, if we compare the old topology and the new topology, apart from the actual node itself going away, the only difference is that the NAME of the pseudonode has changed. All the links and costs to it and from it must be the same as they were before (except for the links to and from the physically failed node of course).
 So if there were some way we could recognize this, we could just
ignore the problem.
        Mike

At 09:33 13/04/2005 +0100, mike shand wrote:
At 11:34 12/04/2005 -0400, Alia Atlas wrote:
In the design team discussion, a question with the Ordered FIBs approach to micro-loop prevention was whether it could handle both "positive" and "negative" news in the same set of topology changes. In general, this was seen as a nice to have.
I believe that the broadcast link case may make this necessary. The
scenario I am thinking of is where the interface from the DR on a
broadcast link fails. As a result, the backup DR will start announcing
the router LSA needed for the pseudo-node. This will look like a
pseudo-node has failed & as if a new pseudo-node has appeared with all
of its links.
Alia,

Yes, good point. That is certainly true, and looks like it could get rather messy:-(
I'm wondering whether it might be possible to deal with this at a higher
level. i.e. rather than consider the pseudonode represented topology
which has been severely shaken about by the DR changeover, is there some
way we can consider the "real" topology which has simply lost a node.
This might involve some special case processing of pseudonodes, but it
might avoid the other complication.
Haven't had a chance to think seriously about this. I just throw it out
as a possibility for further thought.
        Mike



What do others think?

I know that the DT decided on PLSN for the basic approach - but there's the possibility that a different advanced method might become desirable.
Alia


_______________________________________________
Rtgwg mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/rtgwg
_______________________________________________
Rtgwg mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/rtgwg
_______________________________________________
Rtgwg mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/rtgwg


_______________________________________________
Rtgwg mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/rtgwg

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