EASIER: Evolutionary Approach for multi-objective Software archItecturE Refactoring

Abstract

Multi-objective optimization has demonstrated, in the last few years, to be an effective paradigm to tackle different architectural problems, such as service selection, composition and deployment. In particular, multi-objective approaches for searching architectural configurations that optimize quality properties (such as performance, reliability and cost) have been introduced in the last decade. However, a relevant amount of complexity is introduced in this context when performance are considered, often due to expensive iterative generation and solution of performance models. In this paper we introduce EASIER (Evolutionary Approach for multi-objective Software archItecturE Refactoring), that is an approach for applying architecture refactoring based on performance aspects and on the cost of architectural changes. In order to mitigate the complexity related to performance, we exploit the knowledge dwelling in performance antipatterns for more effectively driving the evolutionary algorithm towards optimal solutions. We have implemented our approach on AEmilia ADL, so to carry out performance analysis, antipatterns detection and architecture refactoring within the same environment. We demonstrate the effectiveness and applicability of our approach through a non-trivial experimentation on a case study.