Refining extra-functional property values in hierarchical component models

It is nowadays widely accepted that extra-functional properties (EFPs) are as important as functional properties for system correctness, especially when considering systems such as safety-critical embedded systems. The criticality and resource-constrained nature of these systems necessitate to be able to predict tight and accurate extra-functional property values all along the development, from early estimations to measurements. By using a hierarchical component model that allows implementing components as an assembly of subcomponent instances, the same component can be instantiated in several assemblies, i.e. in different usage contexts. Many EFP values are sensitive to the usage context and knowing information about the enclosing assembly enables refining the values of the properties on the subcomponents. Such refinement is usually not supported and the consistency between refined values and the original ones not ensured. This paper presents the concepts and mechanisms to support EFP refinement in hierarchical component models with explicit property inheritance and refinement policies which formally define consistency constraints between refined value and the original one. These policies are interpreted and ensured for all actors and in all workspaces. The paper also describes the related experiments performed on the ProCom component model.