The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2001-Dec> msg00159



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

CR LDP Local repair

  • From: "Vijayanand C - CTD, Chennai." <vijayc@ctd.hcltech.com>
  • Date: Fri, 14 Dec 2001 14:45:31 +0530

Title: RE: CR LDP Local repair
Don,
If node D receives a new request with same LSPID from B will it return a mapping for that, given that an LSP with same ID exists in D with C as upstream. So , should 'nt the existing CR-LSP be released first.
 
Vijay
-----Original Message-----
From: Don Fedyk [mailto:dwfedyk@nortelnetworks.com]
Sent: Friday, December 14, 2001 1:45 PM
To: Vijayanand C - CTD, Chennai.; mpls@UU.NET
Subject: RE: CR LDP Local repair

Vijay

A couple logic flaws.

1) don't release B-C before new B-D mapping is received.
2) Loose hops shouldn't have bandwidth constraints, think about it, a loose hop has to follow the best route and can't do admission control.

Don


> -----Original Message-----
> From: Vijayanand C - CTD, Chennai. [mailto:vijayc@ctd.hcltech.com]
> Sent: Friday, December 14, 2001 12:48 AM
> To: mpls@UU.NET
> Subject: CR LDP Local repair
>
>
> Hello all,
> I have a question on the signalling funcionality in CR LDP
> for local repair.
>
> Consider the topology below.
>
>                      ---------- C ------------
>                      |                         |
>           --A ---  B ---------------------- D--
>
>
>
> Assume a tunnel exist across A-B-C-D( loose ER Hops as B D  )
>  Suppose a
> route update happens at node B with the next hop for node D
> being node D
> itself(may be due to BW constraints the CSPF has taken this
> decision and LSP
> is now going straight through). At this point node B sends a
> release along
> CD to release the old segment and a request to D.  In case,
> the request
> (which has to travel just one hop) arrives earlier than the
> release( which
> has to be processed at CR LDP at node  C ) at node D, then
> the request would
> be rejected by node D because the request is not a
> mdification request but a
> request for existing LSP itself !!
>
> This situation would occur more consistently if the original
> path that is
> released were more longer than the new path. 
>
> How does CR LDP handle such a signalling scenario?
>
> Thanks in advance,
> Regards,
> Vijay
>
>