Abstract
For modelling and evaluation of communication networks, an analytical description of the stochastic processes is advantageous. In this paper, the approximation method MEDA (Mixed Erlang Distributions for Approximation) is described. It approximates an empirical distribution function with a mixture of two or more Erlang distributions. The first three empirical moments are fitted exactly. Higher moments and other characteristics are taken into account by approximating the shape of the empirical distribution function. The deviation is minimized applying a nonlinear programming algorithm. The successful application of MEDA to an empirical file length distribution is shown. A public domain PASCAL source code is available.

This publication has 2 references indexed in Scilit: