[Tmrg] (limited) measurement of file size vs congestion level

[Lachlan Andrew]

Greetings all,

During the discussion of what traffic parameters to change to test
different levels of congestion, we noted the lack of measurement of
file sizes w.r.t. congestion level.

The measurements reported in Table I of
    K Shah, S Bohacek "High short-term bit-rates from TCP Flows," MASCOTS, 2005
    <http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1521118>
suggest strongly that increased congestion comes from increased mean
file size, rather than simply increased arrival rate.

For different link utilisations, here are the mean file sizes, and
number of flows per hour:
4%:    2.85 kByte    16.8M
5.4%: 4.85 kByte    12.8M
25%:  9.80 kByte     29.2M
35%:  10.1 kByte     39.5M
48%:  13.2 kByte     41.2M

This suggests that to increase congestion by a factor of  x,  we
should increase both the arrival rate and the mean file size by a
factor of  sqrt(x).

Thoughts?

Lachlan

-- 
Lachlan Andrew  Dept of Computer Science, Caltech
1200 E California Blvd, Mail Code 256-80, Pasadena CA 91125, USA
Ph: +1 (626) 395-8820    Fax: +1 (626) 568-3603
http://netlab.caltech.edu/~lachlan