计算机科学与探索 ›› 2011, Vol. 5 ›› Issue (10): 942-952.

• 学术研究 • 上一篇    下一篇

采用AADL的软件系统可靠性建模与评估方法

高金梁, 张 刚, 经小川, 陈 星, 张 辉   

  1. 中国航天工程咨询中心, 北京 100048
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-10-01 发布日期:2011-10-01

Software System Reliability Modeling and Evaluation Using AADL

GAO Jinliang, ZHANG Gang, JING Xiaochuan, CHEN Xing, ZHANG Hui   

  1. China Aerospace Engineering Consultation Center, Beijing 100048, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-10-01 Published:2011-10-01

摘要: 结构分析与设计语言(architecture analysis and design language, AADL)是应用于嵌入式领域的体系结构建模、分析和验证的重要手段。针对系统可靠性随着其规模、复杂度和性能需求的不断提升而愈显突出的问题, 给出了一个基于AADL的软件系统可靠性建模与评估框架:首先建立AADL可靠性模型, 然后将其转换为广义随机Petri网(generalized stochastic Petri net, GSPN)模型后再进行分析, 最后根据分析结果判断是否需要进行模型改进。在研究已有的基本转换规则的基础上, 重点讨论了系统中组件之间错误传播以及表示系统发生模式转换的Guard_Transition属性到GSPN的转换规则。以某飞行控制系统中数据发送和处理单元为实例, 验证了所提转换规则和可靠性建模与评估框架的有效性。

关键词: 结构分析与设计语言(AADL), 错误模型附件(EMA), 广义随机Petri网(GSPN), 转换规则, 可靠性, 模型改进

Abstract: Architecture analysis and design language (AADL) is used in embedded field for architecture modeling, analysis and verification. As for the problem that the reliability of system becomes crucial along with the increasing of its size, complexity and performance requirements, this paper gives an AADL-based software system reliability modeling and evaluation framework: first establishes the AADL reliability model, and then transforms it into gener-alized stochastic Petri net (GSPN) model for analysis, finally judges whether model improvement is needed or not according to the analysis results. Based on the study of the existing basic translation rules, this paper focuses on the transformation rules from error propagations between the system components and the Guard_Transition property representing mode translations to GSPN, and then takes data transmission and process units in a flight control sys-tem for example to verify the effectiveness of transformation rules and framework.

Key words: architecture analysis and design language (AADL), error model annex (EMA), generalized stochastic Petri net (GSPN), transformation rule, reliability, model improvement