Towards System Level Performance Modelling

To manage complexity and to shorten design cycles, industry is forced to look at system level methods for specification and design. These methods focus on the creation of executable system models describing a system in the earliestphases of the design process. They allow correctness properties and performance characteristics to be analysed before the system is actually being realised in terms of hardware and software components. Techniques based on classical process algebras (such as CCS and CSP) and their timed extensions have proven to be very useful to describe complex systems and to verify their functional correctness. These techniques, however, lack the possibilities to express probabilistic behaviour and to analyse performance properties. The reason is that the semantics depends upon the concept of non-determinism of choice where the biases in the resolution of possible transitions are left unspecified. In this paper we make a first attempt into extending a realistic system level modelling language POOSL with the capabilities to express probabilistic behaviour. The extension consists of adding probabilistic information to transitions. It will be shown how the extended language allows the estimation of the performance of a datalink protocol. Keywords— performance modelling, performance evaluation, formal specification, object-oriented methods