Hello all,

I would like to compute the average delay at TCP layer of correct TCP
segments (after retransmissions) knowing the average delay at IP
(without retransmissions) and probability of errors (p). Considering
that the retransmissions do not occurs instantaneously, but after RTO
or triplicate ACK, I guess I should do something like:

delayTCP = delayIP * (1+p*RTO) But how to include the triplicate ACK
effect. Is there any reference that analyzes this issue?