The supportive effect of patterns in architecture decision recovery— A controlled experiment

Abstract

The documentation of software architectural design decisions is important to help people understand the system and the rationale behind architectural solutions. In practice, the documentation of such decisions is regularly done after the fact, or skipped completely. To support software maintenance and evolution, the decisions have to be recovered and described. This is often hindered by the fact that the original architects are not available any more, or they do not completely remember the reasons for making the decisions. Additionally, the whole process is very expensive. In this paper, we hypothesize that architecture decision recovery can be more efficient by focusing on recovering decisions related to applying architecture patterns. To test this hypothesis, we designed a controlled experiment that was conducted to analyze the impact of architecture patterns on the quality and quantity of architecture decisions recovered after the fact. We are able to provide statistical evidence that a focus on patterns significantly increases the quality of decisions, while no conclusive evidence concerning the quantity of decisions was found.