[Tmrg-interest] Interesting Draft Document on "TCP Benchmarking"

Xiaoliang (David) Wei weixl@caltech.edu
Tue, 27 Sep 2005 18:01:41 -0700


Oops. First, I'd thank Cesar for looking at and introducing our work in 
progress. We surely appreciate any comments.

For readers who are interested in the work, there is a slightly better 
organized summary at:
http://www.cs.caltech.edu/~weixl/research/technicals/benchmark/summary.ps
It includes our motivation, work-in-progress and our plan to utilize 
Wan-in-Lab to help the community to run benchmark suites in the future.

The draft pointed out by Cesar contains all the details of our scenarios, 
which are ignored in the summary.:
> http://www.cs.caltech.edu/%7Eweixl/research/technicals/benchmark/draft3.pdf 
> (unfinished draft)
> http://www.cs.caltech.edu/~hegdesan/sigcomtest/   (Some results)

We have been working on this since early 2004. After writing down all the 
scenario details in the original draft, we have been iteratively running 
experiments and simulations on the several high speed variants of TCP, and 
modify the scenario details according to our experiment experience. Since we 
haven't understood some of the results, we haven't published the work yet --  
to minimize the possibility that some of our results may be misleading. 
(e.g. problems of slow loss recoveries in some experiments are probably not 
from TCP algorithms, but from Linux implementation. And some other results 
may be "artefacts" of bugs in certain version of NIC hardware...)

We have a plan to combine these resources into a better organized technical 
report, after we understand the problems carefully. Thanks.

-David

Xiaoliang (David) Wei             Graduate Student in CS@Caltech
http://www.cs.caltech.edu/~weixl
====================================================