Avaliação do desempenho de técnicas de programação orientada a aspectos

Abstract

Aspect-Oriented Programming (AOP) was proposed with the main objective of addressing an important software quality principle that is modularization. The basic idea of the paradigm is to capture crosscutting concerns and handle them independently from the rest of the application. AOP emerged as complement to Object-Oriented Programming (OOP), generating many evaluations and empirical studies about this new paradigm. A variety of metrics were applyed on AOP in order to provide evidence of its benefits or problems, however there is no consensus about the impact on performance of the use of AOP techniques to deal with crosscutting concerns on systems. In order to identify if these techniques related to AOP have impact on system performance, this study proposed a controlled experiment on an academic system of the Federal University of Uberlândia. The experiment had the purpose of assessing factors related to AOP which may influence systems performance. According to the results achieved by this study, factors such as the weaver, number of join points, type of advice and number of lines of code (LOC) affect systems performance when it comes to AOP. This influence, however, showed up to be negligible for common web applications which access databases and have MVC framework. Results also address that, specifically for the load-time weaving process, the stage of loading of classes may affect performance of applications which run on the same environment. These results can be used by architects and developers to reason about system architecture projects concerning the use of AOP.