Hi JP, thanks for the clarifications and suggestions!

Br, Dan. 

From: JP Vasseur [mailto:[email protected]] 
Sent: 17 December 2009 18:17
To: Daniel King
Cc: [email protected]
Subject: Re: [Pce] WG Last Call on I-D Action:draft-ietf-pce-p2mp-req-04.txt

Hi Dan,

On Dec 15, 2009, at 5:51 PM, Daniel King wrote:


Hi JP, All,
 
As a co-editor of the P2MP Solutions
(draft-ietf-pce-pcep-p2mp-extensions-05) I would like to state that the
authors of the P2MP Requirements (draft-ietf-pce-p2mp-req-04) added all the
additional requirements that we identified during the development of the
P2MP solutions document. So I believe that the P2MP Requirements document is
in fairly good shape.
 
We do however have one outstanding issue regarding the solutions document. I
do not believe this directly impacts the P2MP Requirements document itself,
but it’s certainly worth discussing the issue on the list. The P2MP
Requirements draft mentions the requirement for reoptimization of P2MP TE
LSPs:
 
>> 
2.1.10. Reoptimization of P2MP TE LSPs
 
R11: Reoptimization MUST be supported for P2MP TE LSPs as described for P2P
LSPs in [RFC4657].
<< 
 
There is a further request in this section that states "that a P2MP Path
Computation Request SHOULD contain a parameter that allows the PCC to
express a cost-benefit reoptimization threshold for the whole LSP as well as
per destination". This seems like a sensible request. After all, why switch
to a new path if the reoptimization request returns a path that's no better,
or even worse, than the existing path?  
 
We support Objective Functions [RFC5541] in the P2MP Solutions document, and
our solutions allows a path to be computed using a Shortest Path Tree (SPT)
or Minimum Cost Tree (MCT), these will be processed according to the rules
defined in RFC5541, but we do not currently have a "reoptimization
threshold" function. Some of the authors of the P2MP Solution draft think
the reoptimization threshold function is relevant to both P2P and P2MP, and
threshold solution should be applicable to both P2P and P2MP. Therefore the
current P2MP Solutions would state that this requirement will be resolved in
a future document.
 
We would really like to hear comments or suggestions regarding the threshold
reoptimization function. Both from the perspective of a future solution that
supports both P2P and P2MP, and/or this feature should be specifically part
of the existing P2MP Solutions draft.

Good question and for people having long memories, this was discussed almost
5 years ago (!). I had personally advocated for such a solution so I'm
clearly in favor of it and I also believe that the same mechanism should
apply to both P2P and P2MP. Of course, there might be a need for specific
threshold in light of the OF. For example, if we end up defining an OF for
P2MP according which the objective is to minimize the path cost difference
between each leaf and the head-end (to guaranty fairness delivery for mcast
traffic for example), we may need a P2MP specific threshold. So let's try to
make the threshold mechanisms as generic and modular as possible.

Back to your question I would suggest to leave it out from the P2MP solution
draft and have a separate ID covering both the P2P and P2MP threshold
solutions.

Thanks.

JP.


 
Br, Dan/P2MP Solutions Authors.    
 
 
From: [email protected] [mailto:[email protected]] on Behalf Of JP
Vasseur
Sent: 07 December 2009 19:08
To: [email protected]
Subject: [Pce] WG Last Call on I-D Action:draft-ietf-pce-p2mp-req-04.txt
 
Dear WG,
 
Although the discussion on this ID has been relatively moderate, the
document has been stable for a while and discussed at several WG meetings. 
 
This starts a 2-week WG Last Call on I-D
Action:draft-ietf-pce-p2mp-req-04.txt.
 
Please send your comment to the authors and copy the list by Dec 15, noon
ET.
 
Thanks.
 
JP and Julien.


_______________________________________________
Pce mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/pce

Reply via email to