Estimation process of performance constraints during the design of real-time & embedded systems

Abstract

Real Time and Embedded systems always have performance constraints either on response time or on throughput or on device utilisation rate. With conventional design methods these constraints are verified until the testing phase. No estimation of the variables on which the designer has specified some performance constraint is available before. In the frame of the Esprit II project COMPLEMENT2, whose aim is to detect and fulfil the gaps between the different methods and tools that cover the whole design process of a real time and embedded system, an important effort is invested in including the use of performance modelling techniques very early in the development life-cycle in order to compute estimations of the performance criteria. The goal of this paper is to present some of the results obtained around the integration of performance evaluation modelling within design methods: - the taxonomy of performance requirements to classify the types of performance constraints, - the performance information model to bridge the gap between the design entities and the performance models, by introducing annotations to be able to run the derived models in order to provide the efficiency estimation of the system.