Quantifying architectural debts

Abstract

In our prior research, we found that problematic architectural connections can propagate errors. We also found that among multiple files, the architectural connections that violate common design principles strongly correlate with the error-proneness of files. The flawed architectural connections, if not fixed properly and timely, can become debts that accumulate high interest in terms of maintenance costs over time. In this paper, we define architectural debts as clusters of files with problematic architectural connections among them, and their connections incur high maintenance costs over time. Our goal is to 1) precisely identify which and how many files are involved in architectural debts; 2) quantify the penalties of architectural debts in terms of maintenance costs; and 3) model the growth trend of penalties---maintenance costs---that accumulate due to architectural debts. We plan to provide a quantitative model for project managers and stakeholders as a reference in making decisions of whether, when and where to invest in refactoring.