Methods for analysis and design of heterogeneous telecommunication networks

Dorte Møller

Summary

This dissertation is the result of a Ph.D. study initiated by the Institute of Mathematical Modelling (IMM) of the Technical University of Denmark, the Danish State Railways (DSB), and NKT Dedicom. The subject of the study is the performance of heterogeneous communication networks. A heterogeneous network is one that comprises several subnetworks employing different protocols or transmission media. The increasing integration of networks and communication types makes it desirable to develop end-to-end performance models as an aid to network design. The large heterogeneous network of DSB was used as a case study, as one of the main purposes of the study was to assess the usefulness of mathematical modelling in a real-world environment.

The work presented here falls in three main areas. The first area is the study of models for relevant homogeneous subnetworks. Specific subnetworks were chosen for investigation with the case study in mind, namely Ethernet local area networks and buffered communication lines connected by routers. In theis context, a new model was developed of buffered communication links with batch rejection. The model, M-x/PH/1/N, was solved analytically and employed to perform a numerical investigation of the system. Furthermore, a simulation model was used to validate the model.

The second area is the study of communication demands. Here, a theoretical framework for assessing and quantifying demands independently of available services has been developed. The framework has been used to perform an analysis of the communication demands in DSB. Measurements of data traffic performed in DSB have been analyzed and compared to measurements reported in the literature. The third area is the construction of end-to-end performance models of heterogeneous networks. A number of existing models have been surveyed, and a conceptually new approach to generic end-to-end performance modelling has been proposed. The work has resulted in a number of engineering guidelines for network design and a discussion of the feasibility of using stochastic modelling in a commercial organization.

IMM Ph.D Thesis 8, 1994


Last modified August 18, 1996
Finn Kuno Christensen

Back to IMM Homepage