Architecture Smells and Pareto Principle: A Preliminary Empirical Exploration

Architecture smells represent violations of best practices recommended for software architecture that adversely impact various quality attributes of a software system. Though architecture quality is considered very important by the software engineering community, architecture refactoring, given involved high risk and effort, is often avoided by software development teams. In this paper, we empirically explore the properties of architecture smells in the context of the Pareto principle. We investigate the degree of adherence of architecture smell occurrences to the Pareto principle and explore the influence of other related factors i.e., programming language and size of the repositories. To this end, we analyzed 750 Java and 361 C# repositories containing more than 50 million lines of code to detect seven kinds of architecture smells. We found that approximately 45% of the Java repositories follow the Pareto principle. Moreover, C# repositories show significantly higher adherence (66%) to the principle than the repositories written in Java. Our results indicate that the size of the repositories shows a low negative correlation with the Pareto categories. The results imply that software development teams can figure out a few vital components suffering from architecture smells by carrying out the Pareto analysis. It will allow them to optimize their efforts towards making their software architecture quality better.