基于开源软件的有向图研究  

Analyzing Package Dependencies in Open Source Software Using a Directed Random Graph

作　　者：李慧倩[1] 曾大军[1] 郑晓龙[1] 王飞跃[1] 毛文吉[1] 

机构地区：[1]中国科学院自动化研究所复杂系统与智能科学重点实验室,北京100080

出　　处：《复杂系统与复杂性科学》2008年第1期6-13,共8页Complex Systems and Complexity Science

基　　金：国家自然科学基金委基金创新研究群体科学基金(60621001);中科院海外合作伙伴计划项目(2F05N01)

摘　　要：从复杂网络的角度分析大型开源软件内部软件包依赖关系的复杂度和风险情况。以Gentoo Linux系统做为数据源,将系统中的软件包抽象成点,把软件包之间的依赖关系抽象成边,建立随机图,对此进行分析。除应用经典的随机图方法外,还定义了两个基于有向图的新参数:复杂度和风险度分布,用以量化地评估系统的复杂程度与风险性。通过应用两个新参数,观察到Gentoo系统的复杂度和风险度分布经历了一个快速增长的阶段。这种高速增长的原因来自系统中环状依赖结构。In this paper, we study one of the most popular open source Linux distributions called Gentoo Linux under the complex systems analysis framework. We model the packages in Gentoo Linux as vertices and the dependency relationships among these packages as edges. In addition to using common graph measures such as degree distribution, we have developed two additional measures, average complexity and risk distribution, to investigate this graph in an attempt to better capture the complexity of large-size software packages and the potential impact of software defects. These measures are based on modeling package dependency as a directed graph. Our analysis shows that the average complexity and risk distribution of Gentoo Linux has experienced a period of rapid increase. Such increases are closely associated with the circle dependency relationship among some of the heavily depended packages.

关 键 词：开源软件 GENTOO 复杂网络 随机图理论 

分 类 号：N94[自然科学总论—系统科学] TP393[自动化与计算机技术—计算机应用技术]