[Iccrg] MulTFRC update
Michael Welzl
michawe at ifi.uio.no
Tue Jul 6 10:29:22 BST 2010
Hello,
We submitted an update of our MulTFRC draft, in which we tried to
address the
comments that we have received. After it was presented to DCCP, and
there,
it was said that ICCRG would be a better place for the draft, it is
now called
draft-irtf-iccrg-multfrc-00.txt.
http://tools.ietf.org/html/draft-irtf-iccrg-multfrc-00.txt
The majority of the comments revolved around the possible problems with
implementing our equation; we have addressed this by including an
appendix
in which it is discussed that our algorithm provided for this
calculation will
not give underflow, overflow or rounding errors that will adversely
affect the
result of the algorithm.
On our MulTFRC page:
http://heim.ifi.uio.no/~michawe/research/projects/multfrc/index.html
You will now also find two very small Java programs on this page,
which implement the TFRC and MulTFRC equations with a GUI, as
an easy means to test input / output values and obtain reference values
for validating an implementation.
Lachlan commented that the MulTFRC doesn't scale with the
bandwidth*delay
product, which we addressed with a new paragraph under "usage
considerations",
where we document this fact and explain that, for this reason, it is not
recommended to use MulTFRC as a fix to the problem of saturating very
large bandwidth*delay product paths.
It was also requested to have a reference which includes the equation's
derivation. MulTFRC was developed by Dragana Damjanovic in her Ph.D.
thesis,
which is now finished, and available from our MulTFRC page. This
document
includes the derivation of the equation and quite many results:
- validations of the equation using simulations and real-life
measurements
- MulTFRC tests using simulations and real-life measurements
One small mistake that happened to me just now is that I forgot to
explicitly
reference this document in the draft; sorry, I will fix this in the
next update.
I intend to present this update at the upcoming ICCRG meeting.
Any comments are welcome!
Cheers,
Michael
More information about the Iccrg
mailing list