Estimates of distributions of random variables for certain computer communications traffic models

Abstract

A study of multiaccess computer communications has characterized the distributions underlying an elementary model of the user-computer interactive process. The model used is elementary in the sense that many of the random variables that generally are of interest in computer communications studies can be decomposed into the elements of this model. Data were examined from four operational multiaccess systems, and the model is shown to be robust; that is, each of the variables of the model has the same distribution independent of which of the four systems is being examined. It is shown that the gamma distribution can be used to describe each of the continuous variables of the model, and that the geometric distribution can be used to describe the discrete variables. Approximations to the gamma distribution by the exponential distribution are discussed for the systems studied.