Entropy as a Measure of Consistency in Software Architecture

Abstract

In building software architectures, the relations between elements in different diagrams are often overlooked. The first stage of building IT systems is the use of ontology terminology, not software terminology, in the requirements engineering process. Then, when constructing software architecture, IT architects more or less consciously however introduce elements that represent the same classifier on different diagrams with similar names. These connections are called consistency rules and are usually not attached in any way in a modeling tool, and only a significant number of them in the models increase the quality of the software architecture. It is mathematically proved that the application of consistency rules increases the information content of software architecture. Authors show that increasing readability and ordering of software architecture by means of consistency rules have their mathematical rationale. In this article, we found proof of decreasing Shannon entropy while applying consistency rules in the construction of software architecture of IT systems. Therefore, it has been shown that marking selected elements in different diagrams with these same names is, therefore, an implicit way to increase the information content of software architecture while simultaneously improving its orderliness and readability. Moreover, this increase in the quality of the software architecture can be measured by entropy, which allows for checking whether the number of consistency rules is sufficient to compare different architectures, even of different sizes, thanks to entropy normalization, and checking during the development of the software architecture, what is the improvement in its orderliness and readability.

Keywords

• software architecture
• consistency rules
• IT system design
• entropy
• consistency assessment