Applying Search Based Optimization to Software Product Line Architectures: Lessons Learned

摘要

The Product-Line Architecture (PLA) is a fundamental SPL artifact. However, PLA design is a people-intensive and non-trivial task, and to find the best architecture can be formulated as an optimization problem with many objectives. We found several approaches that address search-based design of software architectures by using multi-objective evolutionary algorithms. However, such approaches have not been applied to PLAs. Considering such fact, in this work, we explore the use of these approaches to optimize PLAs. An extension of existing approaches is investigated, which uses specific metrics to evaluate the PLA characteristics. Then, we performed a case study involving one SPL. From the experience acquired during this study, we can relate some lessons learned, which are discussed in this work. Furthermore, the results point out that, in the case of PLAs, it is necessary to use SPL specific measures and evolutionary operators more sensitive to the SPL context.