[e2e] How to increase the flow's rate reliably???

[David G. Andersen]

Jeevandra just mooed:

> I am a research student trying to develop a rate-based transport =
> protocol primarily catered for streaming video on the internet. At the =
> [...]

> However, I've found in my simulations that such a mechanism could induce =
> severe oscillation in the rate particularly when used with rate-halving =
> multiplicative decrease. This is mostly obvious when a huge PacketSize =
> [...]
> 
> I was hoping if anybody could share a light on this and advise me on a =
> more suitable rate increase technique, particularly for streaming. Has =
> [...]

  You want to look at two pieces of research to start out with:

a)  Slowly Responsive Congestion Control Algorithms (Bansal,
    Balakrishnan, Floyd, and Shenker, Sigcomm 2001).

    http://nms.lcs.mit.edu/papers/slowcc-sigcomm01.html

    They examine a generalized class of TCP-friendly congestion
    control algorithms where a less agressive increase/decrease
    mechanism smooths out the wild oscillations that standard
    AIMD gets you.  Parts of this have been implemented in
    the Congestion Manager framework, and since it's a relatively
    simple digression from standard TCP, it should be reasonably
    implementable within other frameworks.

b)  Equation-based congestion control (Floyd et al.)

    http://www.aciri.org/tfrc/

    Part of the goal of TFRC is to increase the stability of
    transfers.  I don't know of an actual implementation of this, 
    but I could well be out of date.  TFRC is more on the experimental
    side of things, and an implementation would likely be more difficult
    beause it diverges further from TCP's mechanisms.

  -Dave

-- 
work: dga at lcs.mit.edu                          me:  dga at pobox.com
      MIT Laboratory for Computer Science           http://www.angio.net/