Implementation and Evaluation of Software Architecture Visualization

Abstract: Background: Software architectures are often large documents containing detailed information about different architectural components, how they interact, the environment in which they operate, and how the components are designed. With such a large amount of data, the sizes of the documents often get larger and it becomes challenging for humans to process in a reasonable time. Visualization is the concept of presenting data in a certain way like a graphical or pictorial format, to make data easier to understand for many people. The concept of visualization is introduced into the software architecture domain to ensure that the lengthy architectural documents are simple to understand. Objectives: The objective of this thesis is to understand the data and relationships in the data and find an accurate visualization technique suitable for the data and implement the visualization as an interactive online tool and evaluate the visualization tool through a comparative user study with an implemented tool against existing practice. Methods: The research method used in this thesis is an Experiment, where the data for the experiment is collected through a comparative user study and collected data is evaluated using subjective and objective metrics. The objective metrics used to evaluate are ‘time’ and ‘correctness’ and the subjective metric is the NASA-TLX assessment form, which the participants are asked to fill out during the user study. The statistic value used to test the hypothesis is calculated using Wilcoxon signed rank test. Results: The results of the experiment are such that the time spent by the users answering the questions in the implemented tool is significantly less compared to the existing method. There is not much difference in the correctness of understanding the software architecture, and the subjective workload experienced by the participants during the user study is significantly less in the implemented tool compared to the existing method. Conclusions: The purpose of this thesis is to create an interactive visualization tool that can effectively display the data and relationships within the data. The tool was evaluated through a comparative user study, using both subjective and objective metrics, which showed significant advantages of using it in terms of time and workload reduction.