[Tmrg-interest] proposed work for the tmrg

[[e-mail] (Sally Floyd)]

In this email I am going to lay out my proposed plan of work for
tmrg.  People are of course welcome to bring up other topics
that are within the tmrg charter.


(1) Metrics for the Evaluation of Congestion Control Mechanisms"

The tmrg has one proposed document, "Metrics for the Evaluation of
Congestion Control Mechanisms", that is on the tmrg web site.
I am planning on doing a round of revisions of this document shortly,
and then I will ask the list for feedback.


(2) Tools for Constructing Scenarios for the Evaluation of Congestion
Control Mechanisms:

The next document that I am planning to start is one on "Tools for
Constructing Scenarios for the Evaluation of Congestion Control Mechanisms".
The plan is to start with the two tools discussed in the 2002 paper
on "Internet Research Needs Better Models", which are tools for
characterizing the aggregate traffic on a link in a simulation,
experiment, or in the real world.  

The two tools discussed in "Internet Research Needs Better Models"
are the distribution of per-packet round-trip times (based on papers
by Jiang and Dovrolis and others in looking at distributions of
round-trip times), and the distribution of per-packet sequence
numbers (based on the years of research by many people on distributions
of connection sizes).

Additional tools that might be of use in evaluating scenarios used
for simulations and experiements, and that would be used for comparing
simulation and experiment scenarios with characteristics of the
real world, include the following.  I would note that the tools
below are all characteristics of an end-to-end path, as opposed to
characteristics of the aggregate traffic on a link.

* Synchronization ratio: the degree of synchronization of loss
events between two TCP flows on the same path.  Relevant to the
convergence times for high-bandwidth TCPs.  Under investigation
by Grenville Armitage and Qiang Fu.

* Drop rates as a function of packet size (e.g., for evaluating VoIP 
congestion control).  Discussed, among other places, in
draft-ietf-dccp-tfrc-voip-02.txt, from
"http://www.ietf.org/internet-drafts/draft-ietf-dccp-tfrc-voip-02.txt".
Relevant for evaluating congestion control for small-packet flows.

* Burst-tolerance (drop rates as a function of burst size).
Relevant to fairness for bursty traffic, and the like.

* Minimizing packet drops (measured as the number of packet drops
that occur in a single loss event at the end of slow-start?).
Relevant to the tradeoffs possible between throughput, delay, and
packet drops.

* And several more, to be described later.


(3) Best Current Practices for Scenarios for Simulation and Experiments:

My hope would be that another item of work for the RG would
be the creation of Best Current Practice sets of scenarios for
simulations and experiments.  This could include some
general scenarios, along with specific sets of scenarios for
evaluating congestion control in high-bandwidth environments,
for congestion control over wireless networks, and the like.
My hope would be that different people would be interested
in being editors for different sets of scenarios, incorporating
feedback from the mailing list.

Note that these would not be benchmark scenarios, with the
promise that if your proposed mechanism worked well in the
benchmark, then it could be assumed to work well in the
global Internet.  They would instead be starting points for
investigating the performance of proposed congestion control
mechanisms.


(4) General information.

I also plan to be a rigorous enforcer of focus on the mailing list,
now that I have gotten my act together, with no tolerance
for random discussions that are not tightly focused on the
main business of the RG.  We shall see how that goes.

TMRG hasn't yet been approved as an official RG of the IRTF,
but it seems likely to me that this will happen soon.  Again.
we shall see.

Regards, and my apologies again for the months of spam on the
mailing list up until now.

- Sally
http://www.icir.org/tmrg/