Thesis Type: Postgraduate
Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Computer Engineering, Turkey
Approval Date: 2018
Student: SINA ENTEKHABI
Supervisor: MEHMET HALİT S. OĞUZTÜZÜN
Abstract:Dynamically reconfigurable systems are able to respond to changes in their operational environments by reconfiguring themselves automatically. Dynamic software product lines are dynamically reconfigurable systems with an explicit variability model that guides the reconfiguration. In this work, feature models are used as the variability model. Features are assumed to be mapped to system's components that realize them. A feature model corresponds to a constraint satisfaction problem (CSP), and determines the valid configurations of the system. An emerging situation in the environment can lead to some relevant changes to the current configuration: some features must be activated, and some must be deactivated. Due to constraint propagation, the status of other features might need to be changed as well. However, considering the feature state migration costs, one would like to avoid such changes to the greatest extent possible in order to mitigate the cost of the disruption to the system's operation. In this work, we devised a dynamic constraint satisfaction algorithm that efficiently considers feature state changes to be applied to the current configuration while confronting changes in the environment so that the new configuration will be valid and the requirements of the new situation will be satisfied with the minimum cost. A set of heuristics regarding feature model relations and the overall structure of feature models are also proposed to enhance the efficiency of the algorithm.