Automated Completion of Partial Configurations as a Diagnosis Task Using FastDiag to Improve Performance

Abstract

The completion of partial configurations might represent an expensive computational task. Existing solutions, such as those which use modern constraint satisfaction solvers, perform a complete search, making them unsuitable on large-scale configurations. In this work, we propose an approach to define the completion of a partial configuration like a diagnosis task to solve it by applying the FastDiag algorithm, an efficient solution for preferred minimal diagnosis (updates) in the analyzed partial configuration. We evaluate our proposed method in the completion of partial configurations of random medium and large-size features models and the completion of partial configurations of a feature model of an adapted version of the Ubuntu Xenial OS. Our experimental analysis shows remarkable improvements in our solution regarding the use of classical CSP-based approaches for the same tasks.