Re: [mpls] Last Call: <draft-ietf-mpls-in-udp-04.txt> (Encapsulating MPLS in UDP) to Proposed Standard

[Xuxiaohu <[e-mail]>]

> -----邮件原件-----
> 发件人: mpls [mailto:[e-mail]] 代表 Curtis Villamizar
> 发送时间: 2014年1月15日 9:14
> 收件人: Wesley Eddy
> 抄送: Scott Brim; Eggert, Lars; IETF discussion list; [e-mail]
> 主题: Re: [mpls] Last Call: <draft-ietf-mpls-in-udp-04.txt> (Encapsulating MPLS
> in UDP) to Proposed Standard
> 
> 
> In message <[e-mail]>
> Wesley Eddy writes:
> 
> > On 1/14/2014 11:06 AM, Eggert, Lars wrote:
> > > Hi,
> > >
> > > On 2014-1-14, at 16:39, Scott Brim <[e-mail]> wrote:
> > >> Lars, I know we're repeating arguments from the last decade. The
> > >> choice is between (1) specifying congestion control around the
> > >> substrate UDP that can be turned off if it causes problems, or (2)
> > >> specifying nothing at this time and adding it later if operators
> > >> want it.
> > >>
> > >> I guess if this can be written as a SHOULD, up to the implementor's
> > >> discretion, then okay.
> > >
> > > I don't think we can leave this up to implementors discretion. We've had
> IETF consensus that Internet communication requires congestion control at least
> since RFC2914. A circuit breaker mechanisms seems straightforward to
> implement.
> > >
> > > As is, I object to this document going forward. The minor benefits of getting
> some better load balancing for MPLS are far outweighed by the risks.
> > >
> > > (I'm also going to shut up now, and let others speak. I think I've
> > > said my bit.)
> > >
> >
> >
> > I'm in basic agreement.
> >
> > Assuming we might all agree that there are conceivable scenarios where
> > a circuit breaker mechanism would be useful, is the real issue that we
> > don't all agree that it could be implemented in a way that's not
> > burdensome and doesn't degrade performance unnecessarily?
> >
> > Or is there still fundamental disagreement about whether the scenarios
> > where the circuit breaker is useful are even valid?
> 
> 
> A circuit breaker (drop all traffic on sign of congestion) at the MPLS over UDP
> level would be the worst thing you could specify and IMO it would guarentee
> zero implementations.
> 
> A circuit breaker might be appropriate for TDM over PW but nothing else.  This
> makes sense because TDM cannot change its bandwidth so the only choice is to
> shut it off.  TDM over PW is an application that can run over MPLS (or GRE or
> L2TP).  The best place to put that circuit breaker (if anywhere at all) is in TDM
> over PW.

I couldn't agree more with Curtis's point. Since RFC3985 has already made a detailed description of congestion considerations about PWs, it seems that the simplest way is just to add a reference to that RFC in this doc, in addition to the text I have proposed in a previous email. The advantage of this choice is: since there is only one place where the congestion control mechanism for MPLS/PW traffic is specified, it's much easier for operators and implementors to refer. In addition, once a more practical congestion control mechanism was found in the future, it only needs to update one RFC, rather than updating many RFCs.

Best regards,
Xiaohu

> Curtis
> _______________________________________________
> mpls mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/mpls