Hi Carles,
This is indeed an important piece of work. The fact that this draft is
maturing in tandem with the evolution of the CoAP-on-TCP darft is really
beneficial for the IoT technology space.
During the last Prague meeting I made some comments towards the end of the
presentation. I take this opportunity to put those comments in the mailing
list in a more organized form. See if you and your co-authors find them
useful.

One thing that I would like to stress upon is that, I would like to see TCP
in IoT as an inheritance of a more generalized class of problem related to
TCP performance for short flows. This is an old problem and has been
studied in many literatures (Example: [1-3]). The case for IoT is a
specialization (the word "specialization" would most likely attribute to
the factors like scalability, h/w constraints, etc.). In [4] one can find a
mathematical definition for short flows for TCP.

(In fact, going by [5], it will not be too wrong to say that IoT is
basically a culmination of different existing technological issues under
one umbrella that predominantly deals with constrained  devices and
networks.)

So, just check if you can deliver the problem statement in a bit
generalized manner if the above makes sense.

Coming to the problem with short flows, the basic problem is the
sub-optimal performance of slow-start and non-availability of enough
duplicate ACKs (dupacks) to start the fast-retransmission. Now , your draft
very rightly takes into account the cases where the window may run over
more than one (and only a few) MSS. While you have mentioned about the
utility of ECN and SACK, probably it would also be useful to mention about
the "limited transmit" algorithm [6]. I do not have readily available
statistics about its implementation in Kernels at present. But, probably it
is available. [6] essentially optimizes on how the fast re-transmit works
for short-flows which do not run over enough segments to ensure sufficient
number of dupacks to indicate a 'softer' congestion and thus prevents the
sender from going into the costly slow-start phase (as RTO remains the only
option to detect congestion in the absence of enough dupacks). Combination
of SACK and [6] may benefit the system. However, I do not have any readily
available study on the performance benchmark for this. But it is an option
worth keeping in this work, I think.

Thank you.
Best wishes for your draft.
------------------------
[1] H. Balakrishnan, et al, “TCP Behavior of a Busy Internet Server:
Analysis and Improvements “, in Proc. Of IEEE Infocomm ’98, CA, USA, March,
1998.
[2] N. Cardwell, et al, “Modeling the Performance of Short TCP
Connections”, Technical Report, University of Washington, October, 1998 (
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.30.2099&rep=rep1&type=pdf
)
[3] K. Avrachenkov, et al, “Differentiation between short and long TCP
flows: predictability of the response time”, INFOCOM 2004
[4] N. Kartik, “TCP optimized for short flows”, Stanford University, June
2003, (http://web.stanford.edu/class/ee384y/projects/download03/nitin3.pdf).
[5] Karen Rose, Scott Eldridge, Lyman Chapin, "THE INTERNETOF THINGS:AN
OVERVIEW", October, 2015.
[6] M. Allman, H. Balakrishnan, S. Floyd, RFC 3042, “Enhancing TCP's loss
recovery using limited transmit” , January, 2001.

On Mon, Oct 16, 2017 at 12:02 AM, <[email protected]> wrote:

>
> A New Internet-Draft is available from the on-line Internet-Drafts
> directories.
> This draft is a work item of the Light-Weight Implementation Guidance WG
> of the IETF.
>
>         Title           : TCP Usage Guidance in the Internet of Things
> (IoT)
>         Authors         : Carles Gomez
>                           Jon Crowcroft
>                           Michael Scharf
>         Filename        : draft-ietf-lwig-tcp-
> constrained-node-networks-01.txt
>         Pages           : 20
>         Date            : 2017-10-15
>
> Abstract:
>    This document provides guidance on how to implement and use the
>    Transmission Control Protocol (TCP) in Constrained-Node Networks
>    (CNNs), which are a characterstic of the Internet of Things (IoT).
>    Such environments require a lightweight TCP implementation and may
>    not make use of optional functionality.  This document explains a
>    number of known and deployed techniques to simplify a TCP stack as
>    well as corresponding tradeoffs.  The objective is to help embedded
>    developers with decisions on which TCP features to use.
>
>
> The IETF datatracker status page for this draft is:
> https://datatracker.ietf.org/doc/draft-ietf-lwig-tcp-
> constrained-node-networks/
>
> There are also htmlized versions available at:
> https://tools.ietf.org/html/draft-ietf-lwig-tcp-
> constrained-node-networks-01
> https://datatracker.ietf.org/doc/html/draft-ietf-lwig-tcp-
> constrained-node-networks-01
>
> A diff from the previous version is available at:
> https://www.ietf.org/rfcdiff?url2=draft-ietf-lwig-tcp-
> constrained-node-networks-01
>
>
> 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.
>
> Internet-Drafts are also available by anonymous FTP at:
> ftp://ftp.ietf.org/internet-drafts/
>
> _______________________________________________
> Lwip mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/lwip
>



-- 
Regards,
Abhijan Bhattacharyya,
*Scientist @ TCS Research, India*
_______________________________________________
Lwip mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lwip

Reply via email to