6280: Performance Evaluation of Communication Networks


HW 3:  Due September 23, 2003 (In Class)




The following questions deal with   the " Modelling TCP Performance Paper"


Practice Problems:


1-  List at least 5 modelling assumptions made in the paper. For each of
the assumptions explain why you believe such an assumption did not affect 
the validity of the results.

2- Show how Equation 20 can be derived from Equation 19

3- Let's assume we would like to take into account the effect of using "slow start"
after a timeout. Sketch how the analysis of the paper may need
to be augmented to account for that.

4- Regarding the validation of the formula:

   a) explain why in figure 7a the TD only model overestimates the throughput
even at low loss probability.

   b) when did they find that the TCP throughput formula derived fails to 
provide a good throughput estimate?

 
Hand-in 

1- In Equation 26 they make the assumption that E(Qhat(W)) = Qhat(E(W)).
To understand when this approximation may be valid in general
describe two scenarios of a random variable W, and a function Q(W).

2-
a) What is the value of B(p) derived in equation 18 when p=1? 
Does this value make sense to you -- why or why not?

b) Apply the fix to equation (7) to match figure 2 and carry through the analysis
and derive the "correct" version of equation (18)

c) Now obtain the value of B(p) of this "correct" version when p=1.
Does this make sense?