Buffer management control in data transport network node

Abstract

The application of analytical queuing theory results in behaviour analysis of a distributed computer network or mobile data system (data transport network). It belongs to the preferred method in comparison to the simulation method. The use of analytical methods allows us to calculate effectively various values of parameters in equilibrium, including the total input intensity of data units to every node of transport network even for the more realistic models than the M/M/1 systems. However, these results are derived assuming an infinite buffer in size at a given node. For practical application, we need to project the concrete number of buffers in every node. This paper describes the method of buffer management control for each decomposed network's node of a data transport network in two real cases. For this purpose, the linear dependence between buffer memory size and input queue size at each node of a data transport network was used. For these two real statistical distributions of incoming data units closed expressions are derived, enabling to calculate the required queue size for both queue size limitation methods (assumption of the unlimited and limited queue sizes). For practical use, a very efficient way of computing queue overflow probabilities was developed.