Comparing several candidate architectures variants : An Industrial Case Study

In system and software engineering, the analysis of architectural variants is most of time irrational and manual. The most common approach for comparing variants is comparing results for each variant evaluation. Most advanced approaches available in architecture evaluation are suffering from three principal weaknesses: the absence of criteria elicitation method, no representation of real-life strategies and no explaination of the outcomes. This paper relates experiments of a MCDA tooled method addressing these weaknesses. The experimentation is supported by an industrial use case consisting in selecting the best platform for an handheld Software-defined Radio. Its architecture description is formalised with model-based design tools. As a result we conclude the experimented approach provides sharper results than classic approach on the class of decision problem exposed by avoiding false positives. This approach seems to be promising to improve the confidence in our Decision Analysis Report and their quality in terms of argumenting the reasons of a decision.