And what do Packet interval times and probabilities represent?
However, the not-so-easy answer has to do with traffic modeling. The most common approach is to define packet traffic in terms of packet intervals (or "inter-arrival") times ? this is the time between packets arriving at a certain point. If the variable 'T' is time and 't' is any measured change in time, then the chance of a packet arriving in any period T+t is P(t), where P(t) is a distribution that describes that probability.
One of the classic probability distributions used is the Poisson distribution (which I won't even try to reproduce here - check a statistics book ? suffice it to say it involves exponentials). It is used because it is mathematically tractable and because it looks about right. However, most finicky network theoreticians agree that it doesn't do the job quite right. Others suggest that Pareto distributions are more apt. No one is quite happy.
No doubt, Comnet III uses some particular probability distribution to define the inter-arrival times of packets. Or has you define it manually.
Tell me if that helps or just makes your head hurt.
This was first published in April 2003