Re: [PWE3] 答复: Re: Flow Label in fat-pw

[Stewart Bryant <stbryant at cisco.com>]

jiang.xiaowei at zte.com.cn wrote:
>
> Sir
>
> Is the flow label some kind of solution for Verizon proposed CTG for
> efficient usage and more control over Trunk so that can bind multiple
> 10G links into one N*10G link instead of installing new 40G/100G system?
I will let Andy answer for himself, but my reference to LAG in the draft
this type of application.
>
> And also, could sequence# be used for ECMP balance? 
Not unless you want to do reassembly of order, which at these speeds is
very difficult. The problem is that the packets are not the same length
and so you will have to buffer then and play them out. This way we split
the flows at source into flows that must themselves be in sequence, but
which can be out of sequence relative to each other. Thus on egress no
storage is needed we simply look at the PW label and switch the packet
immediately out of the PE
>
>
> What's the trade off b/w additional flow label and the sequence #
> reuse/multi-tasking.
I do not understand the question.

Regards

Stewart
>
> Thank you.
>
>
>
> *Stewart Bryant <stbryant at cisco.com>*
> 发件人: pwe3-bounces at ietf.org
>
> 2009-03-27 14:10
> 请答复 给
> stbryant at cisco.com
>
>
> 	
> 收件人
> 	Yaakov Stein <yaakov_s at rad.com>
> 抄送
> 	"pwe3 at ietf.org" <pwe3 at ietf.org>
> 主题
> 	Re: [PWE3] Flow Label in fat-pw
>
>
>
> 	
>
>
>
>
>
> Lucy is correct, we need to turn off s/n or we will have problems.
>
> This is because the s/n is defined per pw not per flow.
>
> Stewart
>
> Yaakov Stein wrote:
> >
> > Lucy
> >
> >
> >
> > Can you explain why you believe there will be misordering ?
> >
> >
> >
> > Y(J)S
> >
> >
> >
> > *From:* pwe3-bounces at ietf.org [mailto:pwe3-bounces at ietf.org] *On
> > Behalf Of *Lucy Yong
> > *Sent:* Friday, March 27, 2009 03:39
> > *To:* pwe3 at ietf.org
> > *Subject:* [PWE3] Flow Label in fat-pw
> >
> >
> >
> > Hi Bryant,
> >
> >
> >
> > One comment about using flow label over a pseudowire is: if the
> > control word is used in a pseudowire and non-zero sequence number is
> > inserted, when using flow labels, it may cause the egress PE burden to
> > perform packet sequencing and increase the possibility of packet out
> > of order. Therefore, it is better to mention in the draft that when
> > the control word presents and non-zero sequence number is used, the
> > flow label should not be used or if it is used, the constant flow
> > label should be used.
> >
> >
> >
> > Regards,
> >
> > Lucy
> >
> >
> >
> > ------------------------------------------------------------------------
> >
> > _______________________________________________
> > pwe3 mailing list
> > pwe3 at ietf.org
> > https://www.ietf.org/mailman/listinfo/pwe3
> >
>
> _______________________________________________
> pwe3 mailing list
> pwe3 at ietf.org
> https://www.ietf.org/mailman/listinfo/pwe3
>