Abstract
The development of feature models enables configurable software systems, however in a collaborative environment different designers may specify and select software and/or hardware features in a conflicting way. Starting from multiple feature models developed by the different designers, we produce some artefacts, such as a merged or a diff feature model representing the commonalities and differences, respectively, to support the designers in achieving an agreement on a shared feature model.
For this last model we compute a partial order of system products based on the inclusion relation to get the maximal and minimal sets of common features. The performance evaluation of the software products is optimized, since bounds are obtained analyzing only the performance of the maximal and minimal elements of the proposed order.
We applied the approach to an illustrative case study and we found that our technique is able to correctly provide performance bounds while minimizing the number of evaluations, thus saving computational time.