RETAkE: Abordagem para teste em tempo de execução de sistemas dinamicamente adaptativos

Abstract

A Dynamically Adaptive System (DAS) supports dynamic runtime adaptations to handle context changes. These adaptations can change the structure or behavior of the system and the logic of its adaptation mechanism. Therefore, one of the main challenges in the area is the execution of verification and validation activities. For instance, considering a system executing in a smartphone, which adapts its features through rules in the condition-action format, it may be necessary to change the rules at runtime to better fit the system to a context as the rapidly depleting of power resources, which may require a reduction in the battery level that activates a functionality. However, this adaptation may insert defects in the DAS, causing it to fail to correctly perform the adaptation. Runtime testing can be performed to verify the system’s adaptation during its operations in the final execution environment. To assist in the execution of these tests, features models can be used, which are high-level representations of the system’s functionalities. However, few work in the literature perform adaptation tests during system execution focusing on adaptation rules. Thus, this master’s thesis proposes RETAkE, an approach for the DAS test at runtime based on the system context and feature modeling. The approach focuses on executing test sequences in the adaptation mechanism, verifying the changed adaptation rules. Additionally, RETAkE also has a property checking technique that supports the testing process. As a secondary contribution, it was implemented a tool that automates the execution of the approach proposed in this work. Moreover, two mobile DASs are used in three different RETAkE evaluations, as follows: a proof of concept for detecting faults, a mutations test, and an analysis of the DASs adaptations in execution time. The tool identified the injected faults in the proof of concept as well as in the mutations testing and the last evaluation showed the impact of the tool during the DASs execution time.