Avaliando o desempenho de sistemas adaptáveis baseados em Aom explorando o Lom

Abstract

Adaptive systems based on AOM can be quite flexible, because you can easily change their behavior due to the level of abstraction of such systems, as entities, relationships and business rules are represented by meta-models. These metamodels should be persisted together with the application metadata in the database, for example. However, developing an AOM system involves a certain level of complexity, but it can be simplified through software reuse, using tools such as the LOM framework, for example. Besides their complexity, another aspect that is always questioned is the performance of such systems, as they demand the constant manipulation of metadata, which implies constant access to databases. So, the research problem in which we focused was the investigation of the performance impact of transforming an object-oriented system, with little focus on adaptability, in an adaptive system using the AOM architectural style in its development. One of the important contributions of this study was the development of part of the LOM framework, which is an alternative to aid the development of AOM systems. The system to manage exercises called EducService was developed in three different versions: a pure object-oriented version and two other ones with the help of LOM in two versions of the persistence layer. The performance evaluation considered three versions of EducService, comparing the results obtained through the execution of predefined scenarios and applying some statistical calculations on the obtained data in order to better analyze them.