Quality-Oriented Design Space Exploration for Component-Based Architectures

Abstract

In this paper we present the DeSiX design space exploration methodology for software component-based systems on multiprocessor architectures. The proposed technique adopts multi-dimensional quality attribute analysis in the early design phases and is based on (1) various types of models for software components, processing nodes, memories and bus links, (2) scenarios of system critical execution, allowing the designer to focus only on relevant static and dynamic system configurations, (3) simulation of tasks automatically reconstructed for each scenario and (4) Pareto curves for identification of optimal architecture alternatives. The feasibility of our methodology is shown by a detailed case study of a car radio navigation system, where te starting point is five candidate system architectures, representing different cost versus performance sensitivity versus reliability tradeoffs.

Authors

Egor I. Bondarev; Michel Chaudron; Peter de With

Published in

Computer Science Reports, TUE Department of Matematics and Computer Science ; February, 2006

back to main page