[e2e] packet-pair probe implementation

[Atsuo J.]

thank you for all the feedbacks.
Rik Wade <rik at rikwade.com> wrote:> I am not aware of any real world implementations of a packet-pair > estimator other than those mentioned. So, even after a few improvement (like packet trains or packet bunches) from original Keshav version, there is still no real world implementation yet. Or, anybody can suggest any implementation, if it is. > However, I did a great deal of work with packet-pair as part of my PhD > thesis in the Real and NS simulators. I have an implementation of > packet-pair for both simulators if anyone would like the source code.Thank you very much. However, if I don't misunderstand, PP is implemented in REAL by Keshav already. Is your PP module in Real different from the original one? What exactly is it? Receiver-side Packet Bunches of Paxson or Packet train of Bengta or other variants? I would be very interested if it's an improved version.  Also, in NS, what exactly is your PP? I think PLM Ns2 module of Legout has provided Receiver-side Packet Bunches already. which variants? Or, is it your own improved version? I have been in your homepage but see no details. If U have done any improvement of PP, it would be very interesting for me.  > At least in a simulation environment packet-pair was very effective, > at least until the congestion became so great that probe packets were > regularly lost.It would be very difficult to show the weakpoint of PP in simulation, I figure. Also, in pathological case that probe packet regularly lost during high congestion, we would need to provide other alternatives to handle situation (just in my opinion).
In my opinion, if we can find a way to measure the available bandwidth correctly, it would simplify the congestion control technique. We can just explicitly adjust the transmission rate accordingly. From several feedbacks, there are a few improvement works on top of PP to measure the available bandwidth properly.  I am still curious that ... Among them, is there any of them matured enough and will be (or has been) implemented in the real world or in the test-bed?  Atsuo

---------------------------------
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.postel.org/pipermail/end2end-interest/attachments/20030503/7d419e73/attachment.html