Analysis And Modeling Of Multi-granularity Software Evolution

The Studies on complexity and evolutionary mechanisms of software system are important issues in software engineering. Applying the classical model used in complex network to analyze software system is lacking of overall structure description with the increasing complexity of software system, and is not suitable for solving corresponding problems in large-scale and complex software systems any more. A model which is more suitable to software network is proposed based on the topological features and macroscopic laws of the software. And it becomes the foundation of the studies on structures and evolutionary mechanisms of software systems.Several open source software as samples are researched. The way of nodes join in the network and the probability of the connection are researched. For the problem of preferential attachment mechanism of BA model which describes inadequately on the connecting direction between nodes in the process of the formation of software network, Varieties of real software systems are empirically investigated. The outcome of the statistics on the directions of new connections is that most of these connections are directed to existing vertices. The statistics on the correlations between the numbers of newly incoming and outgoing connections the existing vertices acquire and their in-degree and out-degree show that the probabilities for existing vertices to acquire new incoming and outgoing connections are respectively proportional to their in-degree and out-degree. And basing on the characteristics of multi-granularity, the software system is abstracted to three granularities. According to the self-similarity in the software network, it gets a statistic of motifs and their frequency of occurrence appearing in a large number of real software networks. Based on this, we propose a multi-granularity software network evolution model. The software network is divided into three levels. Nodes are added to the network in the units of scale-Ⅱ and scale-Ⅲ grains. The directed preferential attachment mechanism of software networks are used when a grain join into the network. We compare the network generated by the model with the real software network and the network generated by BA model in many aspects; verify the validity of the model. it show that the model is better than BA model. Finally it discusses the impact of the parameters in the model on the generated network and the trade-off between cohesion and coupling, reuse and modularity in software system.This work is useful for better understanding the structure of software system and the process of the formation of software network, and has an import role on guiding the iterative development of software system.