The MPLS WG Archive

[Curtis Villamizar <curtis@workhorse.fictitious.org>]

In message <BCFB7F5FCA46D3119EE10050048279E087E74E@nt_d2300.chromisys.com>, Joh
n Drake writes:
> Curtis,
> 
> The intent of the Notify is to keep everything undisturbed while restoration
> occurs.  In your example, a failure of a lambda  would be communicated to
> the lambda LSP endpoints with a Notify.  The endpoints could then switch
> over to an alternate path, if M:N path protection was being used, or else
> establish a new LSP.  One of the other benefits of Notify is that since the
> original LSP has not been torn down, segments of it can be reused for the
> new LSP via the RSVP make before break capability.

The point is that the outermost LSP in a heirarchy of LSPs should be
rerouted or otherwise restored to service.  If this is a LSC that uses
M:N optical path protection, then just switch to another lambda and
affect only the LSC, not the PSC inside it.  Nothing magic about Notify.

> Also, we will eventually be doing multi-area TE, and relying on IGP flooding
> will not work in that context.

Neither will LSP setup except using loose hops.  If you broke the IGP
into areas it was too big to scale well.  If it is that big, then the
LSP from outside the area should ride within a PSC or other tunnel.
Then the restoration occurs within the area only not affecting any
node outside the area that is serving as ingress for any tunnel that
goes through the area.

> Thanks,
> 
> John

Still unconvinced.

Curtis