Software library for analysing M/G/1-type systems and its application in computer network performance evaluation

摘要

Analytical models are an important tool in providing a quantitative analysis of computer and communication systems. Matrix analytic techniques provide exact solutions for a general class of stochastic models named M/G/1-type models. These models can be used to study, for example, network switches with correlated arrival traffic and/or complex scheduling algorithms. However, the modelling effort using this technique can be difficult and time-consuming. The modelling effort can be drastically reduced by using appropriate software tools. The authors present the software library MG1TYPE for the analysis of M/G/1-type Markov chains. MG1TYPE is general, robust and computationally efficient. It provides a range of performance measurement figures including the distribution of the steady-state probabilities and its moments. Applications of MG1TYPE in the analysis of relevant computer communication problems are presented.