Re: [ippm] About Path Congestion Concept, welcome to discuss//Re: About path congestion metric, welcome to discuss

[Jose Ignacio Alvarez-Hamelin <[e-mail]>]

Dear Joanna,

I have read your proposition, and I work on Internet measurements for long time, I have some questions and comments. 

First, as Al Morton already pointed out, there is our work on the delay measurements (https://tools.ietf.org/html/draft-ietf-ippm-route-04)  which is a first step to measure congestion. 

I would like to have a clear definition of congestion, which is, even in the specialized bibliography, a rough and variable concept. 

Then, do you really tried to measure paths in the proposed way? If you perform this in the Internet you will find several things: first, if are measure average RTTs the results will largely varies along the time window and also in two consecutive windows; second, for a fixed window and two consecutive hops your dpr = I3 -I2 could be negative. Even though using other statistics (median, percentiles) you will find odds results. 

If you take attention to our draft AURA, we work carefully to obtain something with certain statistic significance and sense. The problem is that traffic is non-stationary and have a short and long range correlations, yielding on a heavy tailed distribution; which in turn is not easy at all to treat it. This is the reason why you cannot just make a couple of measurements and get something significative in the Internet. My research group is sending a work to IMC 2019 on the congestion analysis, where we deal with real and large links in the Internet.

Best wishes, 

J. Ignacio Alvarez-Hamelin 
http://cnet.fi.uba.ar/en/

===================8<-----------------------
Enviado desde mi CarroT

> On 15 Mar 2019, at 00:26, Dangjuanna <[e-mail]> wrote:
> 
> Hi everyone,
> In the past few days, I learned about the hot spots of discussion on the mailing list. I feel that everyone is currently keen on the definition of the field that the message carries, but the test method is so equally important.
> I think path congestion and increased network traffic are the most realistic and grounded issues. I hope that everyone will pay more attention to the resolution of some practical problems.
> For example, path quality detection, especially path congestion.
>  
> 1.       My multipath concurrent measurements can support not only E2E measurements, but also hop-by-hop data feeds. In other words, it can be seamlessly compatible with popular protocols such as IOAM or Postcard mode.
> draft-dang-ippm-multiple-path-measurement-01.txt (https://tools.ietf.org/id/draft-dang-ippm-multiple-path-measurement-01.txt).
> 2.       My congestion test method uses not only the delay measurement, but also the throughput of the path.
> draft-dang-ippm-congestion(https://tools.ietf.org/html/draft-dang-ippm-congestion-01)
> 3.       Regarding the congestion measurement of delay, I think the method of one-way delay measurement is optimal. This method does not support clock synchronization, which I think is very important.
> 4.       I think its principle of RTT-like measurements determines its accuracy is not enough. Moreover, the PING/TRACEROUTE method requires the CPU or NP to process, and my method can be processed directly on the data surface, so the accuracy and timeliness are stronger.
>  
> Regards,
> Joanna
>  
> From: Dangjuanna 
> Sent: 2019年3月11日 19:56
> To: 'MORTON, ALFRED C (AL)' <[e-mail]>; 'ippm@ietf..org' <[e-mail]>
> Subject: RE: About Path Congestion Concept, welcome to discuss//Re: About path congestion metric, welcome to discuss
>  
> Dear Morton,
> I’m very happy to discuss with you.
> I have read “AURA” WG draft carefully and made the detailed notes.
> Certainly, We are all concerned about the measurement of the path. Especially  I am more concerned with the measurement indicators and measurement methods of path congestion.
> “AURA” WG draft describe there is path congestion requirement.  Just right, I provided the idea of congestion measurement.
> Do you agree with me? I guess you should be very interested at my approach.
> I am looking forward your reply.
>  
> Regards,
> Joanna
>  
> From: Dangjuanna 
> Sent: 2019年3月11日 9:16
> To: 'MORTON, ALFRED C (AL)' <[e-mail]>; [e-mail]
> Subject: RE: About Path Congestion Concept, welcome to discuss//Re: About path congestion metric, welcome to discuss
>  
> Dear Morton，
> Thank you for your reply.
> Last weekend I have looked through your draft named draft-ietf-ippm-route-03.  I value your thoughts very much.
> Indeed, some focus of our thinking is on the same place. On this point, I am very happy that you are also concerned about this point. My current focus is mainly on path congestion, in the scenario of equal-cost multi-path (ECMP) mode and unequal-cost multiple (UCMP) mode.. First, I consider how to measure path congestion.
> Next I will read the draft you suggested.  I will ask you in time if I have any questions.
> If you have any good ideas or corrections, you are welcome to discuss more.
>  
> Best wishes,
> Joanna
>  
> From: MORTON, ALFRED C (AL) [mailto:[e-mail]] 
> Sent: 2019年3月8日 20:58
> To: Dangjuanna <[e-mail]>; [e-mail]
> Subject: RE: About Path Congestion Concept, welcome to discuss//Re: About path congestion metric, welcome to discuss
>  
> Hi Joanna,
>  
> Your proposal sounds similar to parts of the current WG draft:
> https://tools.ietf.org/html/draft-ietf-ippm-route-03
>  
> particularly section 5 and higher (where I would relate
> delay to congestion), although discovering
> the Path Ensemble is the first step (earlier sections).
>  
> Please take a look at the “AURA” WG draft and comment.
>  
> thanks,
> Al
> (for the AURA co-authors)
>  
> From: ippm [mailto:[e-mail]] On Behalf Of Dangjuanna
> Sent: Wednesday, March 6, 2019 2:31 AM
> To: [e-mail]
> Subject: [ippm] About Path Congestion Concept, welcome to discuss//Re: About path congestion metric, welcome to discuss
>  
> Hi  everyone,
> Recently I have submitted the draft named draft-dang-ippm-congestion.
> There are some questions I need to discuss with the experts.
>  
> Q1: Is path congestion required? Is this requirement strong?
>   Import ServiceA Traffic to the path1
>                   \
>                    \      
>                    NodeA----------NodeN1----------NodeN2----------NodeB
>                                  Figure: Path1
> As figure Path1 is shown,  
> [Premise] Path1 is from Node A to NodeB. Path1 may be MPLS TE, SR TE Tunnel or VXLAN.
> [Requirement] Get the congestion degree of Path1 in order to make sure the service experience. 
>  
> Before, everyone was very concerned about the congestion of single device nodes. Along with the network development，we should pay more attention to the business perspective. Therefore, the path perspective of carrying the service is also required. I understand that the requirement for path congestion exists.
>  
> And I think this demand will become more and more important.
>  
> Q2: What is path congestion?
> For the sake of explanation, I introduce a concept named Path Queue. The PathA Queue is activated if congestion occurs at any node along the Path1. Furthermore, when the PathA Queue is activated, it indicates that the path has been congested.
>  
>  
> What do you think of it?
> I am very happy and open to get everyone's thoughts.
>  
> Thank you,
> Joanna
>  
> 发件人: Dangjuanna 
> 发送时间: 2019年3月4日 17:38
> 收件人: '[e-mail]' <[e-mail]>
> 主题: About path congestion metric, welcome to discuss
>  
> Hi everyone,
> I have submit a new individual draft named A One-Path Congestion Metric for IPPM.
> Welcome to discuss.
> Best wishes,
>  
> Joanna Dang
> ===============================================
> A new version of I-D, draft-dang-ippm-congestion-00.txt has been successfully submitted by Joanna Dang and posted to the IETF repository.
>  
> Name:               draft-dang-ippm-congestion
> Revision:  00
> Title:                  A One-Path Congestion Metric for IPPM
> Document date:       2019-03-04
> Group:               Individual Submission
> Pages:               10
> URL:            https://www.ietf.org/internet-drafts/draft-dang-ippm-congestion-00.txt
> Status:         https://datatracker.ietf.org/doc/draft-dang-ippm-congestion/
> Htmlized:       https://tools.ietf.org/html/draft-dang-ippm-congestion-00
> Htmlized:       https://datatracker.ietf.org/doc/html/draft-dang-ippm-congestion
>  
>  
> Abstract:
>    This memo defines a metric for one path congestion across Internet
>    paths.  The traditional mode evaluates network congestion based on
>    the bandwidth utilization of the link.  However, there is a lack of
>    E2E path congestion that is truly service oriented.  So A Path
>    Congestion Metric is required.
>  
>                                                                                  
>  
>  
> Please note that it may take a couple of minutes from the time of submission until the htmlized version and diff are available at tools.ietf.org.
>  
> The IETF Secretariat
>  
>  
> _______________________________________________
> ippm mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/ippm