A Generic Macroscopic Topology of Software Networks - A Quantitative Evaluation

Abstract

The dependence among modules in a software system usually is represented as a network, in which the nodes are the modules, and the edges are the connections between the modules. The relationships among modules in software systems are hard to assess, especially in large programs. Knowing the nature of the software system structures is very important to improve maintenance tasks and other challenging tasks in software development. A previous work of the authors of this paper has defined a model to the topology of software networks, named Little House. This model is a generic macroscopic view of software systems, and it is an adaptation of the well-known Bow-tie model. According to Little House, a software network can be partitioned into six components, in such a way there is a special pattern of connections among them. This paper describes the results of a quantitative evaluation of Little House. The aim of this work is to investigate whether the components of Little House can be described by any pattern of software metric values. The results of this evaluation indicate that in the software systems developed currently there are two main components of Little House that have critical values of metrics. This finding suggests that classes from those components should be carefully considered when maintenance tasks are performed in the program.