Modelling and performance analysis of telecommunication systems (EES.5202)

This proposal responds to the industrial need for methods and tools for system-level architecture design exploration of complex telecommunication systems. The focus in this project is developing techniques for performance modelling and performance evaluation based on system-level behaviour models. These techniques are necessary to handle both current complexity of system design and network traffic management. Dimensioning of buffers (queues), bandwidth of channels, and the distribution of traffic with the required quality of service over the communication resources and computing resources of telecommunication systems become a key issue (technically and commercially). Although the research is carried out in the area of telecommunication, the resulting techniques are not problem-specific and can be used in different application areas to model, analyse and design complex systems at a high level of abstraction. Examples of systems are telecommunication systems, multimedia systems, industrial control systems, consumer electronics, and distributed embedded systems in general. The need for analysis of system-level performance is a common factor in the design of many of these systems.

The focus in this project is on the development of theories, methods and tools for performance analysis enabling the evaluation and comparison of alternative candidate system and network architectures. To this end the concepts required for system-level modelling and the quantitative properties that play a role in system-level design will be classified. Probabilistic extensions of timed process algebras are required that allow the expression of the concepts and the analysis of the properties. The theory on extended algebras will be built in the industrial-strength modelling language POOSL and the accompanying software tools will be extended accordingly. The language and tools, developed within the TUE, have already been applied successfully for system-level design and qualitative (correctness) property analysis in a number of industrial cases. Once this language is able to capture probabilistic information, quantitative (performance) properties can be deduced as well by interpreting models as Markov chains. Both analytical and empirical techniques for extracting performance figures will be developed in this project.

Five top industrial research laboratories in the telecommunication area (KPN Research, TNO-FEL, Lucent Technologies, Alcatel and IBM Research) have acknowledged the need for this research. They agreed to participate in a utilisation group. They will supply pilot projects, actively participate in these projects and they will provide the infrastructure that is not available within the TUE. They recognise that the proposed interdisciplinary co-operation between experts in telecommunication systems design and experts on formal methods is a prerequisite for the necessary breakthroughs on performance analysis. All partners stressed the need for this analysis and the (financial) risks that emerge from a lack of adequate tools.

For results, see http://www.ics.ele.tue.nl/~poosl.

Five companies and one other university are involved in this project.