Research On Architectural Degeneration Of Software System From Defect Analysis Perspective

In order to meet the requirement of changes in demand, software has to evolve constantly. As software evolves over time, the architecture of a software system will inevitably change, and it will gradually lead software architecture to degenerate. In turn, architectural degeneration will result in the difficulties of software evolution, or even an end of the software life cycle. So, as early as possible to evaluate or diagnosis architectural degeneration, are of great significance for steady software quality and prolong the life of software.This paper will observe architectural evolution from the defect perspective, emphatically focused on architectural degeneration.Firstly, this paper gives a comprehensive introduction to the studies of software defect, including the concept of defect, various technologies of defect classification and analysis, and some important experience conclusions. Then the relationship between defect and architectural degeneration is discussed:architectural degeneration can be manifested by defect; and there are some key factors in architectural degeneration that are degeneration-critical components and degeneration-critical relationships.To obtain requisite defect data for architecture analysis, this paper proposes a new method for defect classification, named DC4DA. A multidimensional data model is used, software development process is considered, and the method can effectively reduce redundant and useless information, it has high flexibility as well. The next, we present two kinds of defect metrics which have combined the severity of defect, and based on them, we proposes a model of defect architecture. The proposed model has a hierarchical structure, which clearly shows the defect situation of a software release from different hierarchies, and identifies the critical degenerate structure. Then through cumulative release analysis, we confirm the degenerate structure based on their persistence of degeneration as the basis of optimization, reconstruction or reorganization. Finally, the feasibility of the presented method and model is verified by a prototype system, and the feasibility and effectiveness of the two presented defect metrics are proved through an instance.