The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2001-Oct> msg00200



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

draft-gan-fast-reroute

  • From: francis.arts@alcatel.be
  • Date: Fri, 12 Oct 2001 07:20:44 +0200
  • Cc: Arthi Ayyangar <ayyangar@apollo.mctr.umbc.edu>, mpls@UU.NET
  • X-MIMETrack: Serialize by Router on BEMAIL06/BE/ALCATEL(Release 5.0.8 |June 18, 2001) at10/12/2001 07:20:59,Serialize complete at 10/12/2001 07:20:59


Arthi, Alias,

* When reading the gan-draft, I assumed that node 2 would send [@(2-1), @(3-2), @(4-3), @(egress-4)]. I deduced this from the statement on p.10 (2nd paragraph) of the draft saying that <<the branching node MUST not mix the messages for the main and the detour LSP's>>.


* A similar question as for the RRO shows up for the path MTU: Which path MTU does node 2 send in the RESV to node 1 after the detour [node 2 - node 4] has taken over: The path MTU of [node 1 - node 2 - node 3 - node 4 - egress] or the path MTU of [node 1 - node 2 - detour (node 2, node 4) - egress]?

In order to avoid fragmentation in the middle of the lsp, it is preferable to pass the path MTU of [node 1 - node 2 - detour (node 2, node 4) - egress].



In order to answer all these questions (and new similar questions that may show up), we need to look at the general question: How do the signalling contexts of the main LSP and the detour LSP interact once the detour LSP has taken over?

Regards,

        Francis.

-------------------



To: Arthi Ayyangar <ayyangar@apollo.mctr.umbc.edu>
cc: Francis ARTS/BE/ALCATEL@ALCATEL, mpls@UU.NET
Subject: Re: draft-gan-fast-reroute




At 06:40 PM 10/11/01 -0400, Arthi Ayyangar wrote:
>Hi,
>
>>
>> I have a question on the RRO handling in <draft-gan-fast-reroute>.
>>
>>        Node1 ====> Node2 ======> Node3 =====> Node4 ======> Egress
>>                       v                       ^
>>                       |                       |
>>                       +-----------------------+
>>
>> Notation: @(3-2) == IP address of node 3 on the link [node 3 - node 2].
>>
>> Assume that the link between node 2 and node 3 fails. Then the detour
>> [node 2 - node 4] is taking over. But what does node 2 then send in the
>> RRO object of the RESV message to node 1: Does node 2 continue to send
>> [@(2-1), @(3-2), @(4-3), @(egress-4)] or does node 2 send [@(2-1),
>> @(detour between 4-2), @(egress-4)] or ...?
>>
>---> It would send , [@(2-1), @(detour between 4-2), @(egress-4)]
>as the RRO in the Resv once the detour has taken over.
No, the RRO in the RESV for the Primary LSP would not change.  Node 2 would
continue to send the same RRO as before the failure.
If you look at lsp-tunnel-09.txt, there is a "local protection in-use" flag
which could be set in the IPv4 subobject of the RRO, but the implementation
of draft-gan with which I am familiar doesn't do this.
Alia


GIF image