Otimização de arquitetura de software utilizando sistema de colônia de formigas

Abstract

Architectural design is a critical phase of system development, because decisions taken in this phase have an important impact over the cost and quality of final system. One of the biggest problems faced by software architecture is the high cost of changes and the high dependency on the software engineer. To solve this problem, architectural optimization methods have been used to propose guidelines and recommendations to identify architectural elements, recover and optimize architectures. Objective: In order to avoid architectural changes during development and to make this activity less dependent of the software architect, this work investigates the utilization of ant colony methaheuristic to optimizate the architecture. Method: Firstly, the use of ACO to architectural optimization problems requires use to define the type of architecture accepted by the algorithm, which was component based architecture for our study. Then, we define a representation model of architecture. Finally, we determined the architectural quality metric used to evaluate the solution and the metric to analyze the architectural style. Results: The proposed method was evaluated using experiments with an existent software, observing if the architecture optimization was satisfactory. Considering the adopted quality metric, Modularization Quality (MQ), ACO achieved satisfactory results, generating solutions with higher MQ values than original architecture values. Conclusions: The proposed method proved efficient to find good solutions considering the adopted quality metric.