RE: [PWE3] PWE3 new WG items?

[benjamin.niven-jenkins at bt.com]

So what you are proposing is the problem of signalling the QoS parameters that should be associated with a given PW.  As PW signalling is based on MPLS signalling, I believe that the signalling of QoS parameters is a generic MPLS problem and not one specific to PWs.  In which case it is a problem that should be addressed by the MPLS WG and not the PWE WG.

Ben

-----Original Message-----
From:	Shah, Himanshu [mailto:hshah@ciena.com]
Sent:	Tue 17-Feb-04 19:21
To:	Niven-Jenkins,B,Ben,XGH5 R; pwe3@ietf.org
Cc:	
Subject:	RE: [PWE3] PWE3 new WG items?
Hi Ben,

Lets suppose PE1 and PE2 has PW between them.
When exchanging PW-FEC, PE2 tells PE1 what PW-Label
to use when sending traffic from PE1 to PE2.

With QOS, PE2 also tells PE1, what QOS to use when
sending traffic from PE1 to PE2. In a traffic engineered
mpls core, PE1 would then select a (CAC enabled)transport tunnel
(assuming there is more than one) that suits PW's QOS 
requirement that goes from PE1 to PE2. It is also possible to
create and/or adjust transport tunnel aperture that suits PW's 
QOS requirement.

This is the motivation.

/himanshu



> -----Original Message-----
> From: benjamin.niven-jenkins@bt.com
> [mailto:benjamin.niven-jenkins@bt.com]
> Sent: Tuesday, February 17, 2004 1:44 PM
> To: pwe3@ietf.org
> Subject: RE: [PWE3] PWE3 new WG items?
> 
> 
> Folks,
> 
> > QOS stuff of course... :-)
> > 
> > We submitted PWE3 Control protocol extension to 
> > exchange capabilities as a way to add more stuff
> > and to negotiate. I think QOS for PW is important.
> 
> An observation/fact: A PW acts as a client to a server layer 
> (be it MPLS, IP or whatever) and as a result the QoS that can 
> be provided by a PW cannot be better than the QoS provided by 
> the server layer used by the PW.
> 
> My comments:
> 1) Assuming that the QoS in the available server layers (e.g 
> EXP bits/DiffServ for the MPLS/IP server layers) is adequate 
> (and it could be argued that it isn't) then given that a PW 
> is basically a MPLS LSP with sequence number support, why 
> does PW QoS need its own specification?  Why can't a PW use 
> the same QoS mechanisms as MPLS LSPs?
> 
> For example, why would a service provider deploying PWs over 
> a MPLS core want to use two different but interacting QoS 
> mechanisms (one for PWs, one for MPLS LSPs) and add 
> significant additional fault conditions/complexity/opex to 
> their network when they could use the same QoS mechanism for 
> both PWs and MPLS?
> 
> 2) Assuming that QoS in the available server layers is not 
> adequate then seeing as PW QoS is dependant on the server 
> layer QoS, shouldn't the problem of specifying adequate QoS 
> in the available server layers be addressed first?  Once 
> there is adequate QoS in the available server layers then 
> comment (1) above applies.
> 
> So either:
> QoS for MPLS is adequate and ergo the problem of QoS for PWs 
> has already been addressed.
> 
> Or, QoS for MPLS is inadequate and ergo QoS for MPLS should 
> be addressed first and then the problem of QoS for PWs has 
> already been addressed.
> 
> Ben
> 
> The above is my personal view and does not necessarily 
> reflect the opinion of my employer.
> 
> _______________________________________________
> pwe3 mailing list
> pwe3@ietf.org
> https://www1.ietf.org/mailman/listinfo/pwe3
> 




_______________________________________________
pwe3 mailing list
pwe3@ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3