Um Método de Adaptação para Modelos de Features de Linha de Produtos de Software Dinâmica.

Abstract

Software Product Line (SPL) is a paradigm used in the management of features common to a set of complex systems, which helps in managing variability, decreasing costs and time of implementation and increasing support for maintainability. In an SPL the management of variability is done in a static way with the product configurations being chosen at development time. From the moment that systems start to be part of people’s daily lives, there is a need for these software to run continuously, adapting to changes in the environment and user needs. For the management of the variability of these new systems, the concept of Dynamic Software Product Line (DSPL) emerged. The representation of an DSPL can be done using the feature model (FM) notation. One of the main challenges of an DSPL is the management of FM configurations at runtime, detecting changes in context from a set of adaptation scenarios and varying between possible configurations. In this context, this work aims to develop an adaptation mechanism for DSPL FM that allows the modeling of agents and contexts, and to support dynamic variability. The mechanism transforms the restrictions into rules that condition the activation of each feature, the contexts are also noted in the corresponding features to be activated when changes occur. The development of the mechanism used the concepts of the MAPE-K model and was implemented as an extension of the DyMMer web tool. To evaluate the mechanism, a performance test was carried out with models of simple and complex features and a proof of concept with an expert. The results of the performance evaluation revealed that the mechanism performs the adaptation in a great time, losing a little performance as the FM becomes more complex, but still remaining within an acceptable limit. The results of the evaluation with the specialist showed that the mechanism helps in the modeling of agents and adaptation contexts, as well as intuitively presenting the reconfiguration process at runtime. Thus, it was found that the adaptation mechanism well buys the role of modeling the agents and contexts of an DSPL FM and the execution of the dynamic reconfiguration process.