大规模软件宏观拓扑结构度分布及其演化分析

Analysis of Degree Distribution and Its Evolution of Large-Scale Software Macro-topology

基于大量开源软件源代码度量数据,根据大规模软件宏观拓扑结构体现出来的复杂网络特征,将软件结构抽象为网络拓扑.在这基础上,采用复杂网络的度量分析方法,分别从无向图网络的度分布和有向图网络的出入度分布两方面讨论了大规模软件宏观拓扑结构度分布所体现出的无尺度特征.认为随着软件演化,度分布系数和入度分布系数呈下降趋势,最大节点入度值则呈增大的趋势;而出度分布系数和最大节点出度值则变化有限.然后结合软件工程实践探讨了上述现象形成原因,认为软件系统其结构还有进一步优化的空间.

Based on the metrics data acquired from a large number of source codes of open source softwares and the complex network characteristics embodied by large-scale software macrotopology, the software architectures were abstracted to be a network topology. Then, by the metrical analysis of complex network, the scale-free characteristic embodied by degree distribution in large-scale software macro-topology was discussed in two ways, i. e., the pdf-degree distribution in undirected graph and the pdf-degree distributions of in-degree and out-degree in directed graph. With the evolving software, both the degree distribution coefficient and in-degree distribution coefficient tend to decrease but the in-degree of maximum node tends to increase. However, the fluctuation of out-degree distribution coefficient and out-degree of maximum node are not obvious. The reasons why the degree distribution thus varies are discussed in association with the software engineering in practice, and it is therefore concluded that there is room for further optimization in the architecture of software system.