[e2e] TCP un-friendly congestion control

Steven Low slow at caltech.edu
Fri Jun 6 00:48:08 PDT 2003


Hi Michael

You can stabilize TCP either with or without active queue management
(within the context of fluid model).   There is a class of TCP/AQM
combinations that can maintain linear stability (cf papers by Paganini,
Srikant, and Vinnicombe, Misra, etc. as well as ours).   In all these
schemes, linear stability has been shown for general networks, with
multiple links and heterogenous RTTs.

The FAST algorithm we implemented and tested in live networks
does NOT require AQM, i.e., it assumes DropTail routers.  
We will have a paper that describes
the algorithm and its implementation in a month or so.

Steven



Michael Welzl wrote:

>Hi,
>
>
>>For the past couple of months, whenever someone has bounced up and
>>down extolling the virtues of Katabi's et al.'s XCP and saying
>>something should be done about it, I'd say something like 'gee, that's
>>not the only new(ish) kid on the block. What about, oh, Caltech's FAST
>>work? Which doesn't require the same degree of midpoint modification?
>>And has actual not-just-in-ns implementations? No. N-S.'
>>
>
>Now I won't claim to understand all the maths behind Steven Low's
>work either - but, as far as I understood it, it does at least
>require one midpoint modification: active queue management.
>I think that's how stability is ensured.
>
>I wonder if it's possible to create a TCP-like mechanism that
>would be stable (assuming a fluid model) with no tailored
>active queue management  ... in particular, I wonder if this
>could be shown in the heterogeneous RTT case.
>
>Best regards,
>Michael
>
>

-- 
_____________________________________________
Steven Low                assoc prof, cs & ee
netlab.caltech.edu        tel: (626) 395-6767
slow at caltech.edu          fax: (626) 568-3603







More information about the end2end-interest mailing list