分布式实时系统的软件故障注入  

Software implemented fault injection for distributed real-time systems

作　　者：徐光侠[1,2] 陈蜀宇[3] 常光辉[1] 刘宴兵[2] 刘国良[1] 

机构地区：[1]重庆大学计算机学院,重庆400044 [2]重庆邮电大学软件学院,重庆400067 [3]重庆大学软件学院,重庆400044

出　　处：《重庆大学学报（自然科学版）》2010年第2期62-68,共7页Journal of Chongqing University

基　　金：教育部新世纪优秀人才支持计划(NCET-04-0843);科技部国际科技合作项目(2007DFR10420);国家自然科学基金资助项目(60973160);重庆市自然科学基金资助项目(CSTC;2008BB2307)

摘　　要：针对分布式实时系统的可信验证的难题,建立通用故障模型,将故障模型分为:内存故障、CPU故障、通信故障和服务故障4种。提出一种建立在分布式实时系统环境中的软件故障注入系统结构,它分为3个层次:目标系统、通信网络、软件故障注入系统。软件故障注入系统分3个部分进行设计,软件故障注入器、数据收集模块和故障数据分析模块。对基于该结构的软件故障注入过程进行了说明,实现软件故障注入系统并做了相应的实验分析,实验检测到故障多数为通信故障、内存故障和CPU故障,其覆盖率分别为37.68%、15.47%和15.17%。实验证明这种体系结构很适合分布式实时环境的应用,同时也为进一步研究软件可信验证提供了理论基础和实例依据。Aiming at the problem of dependability validation in the distributed real-time systems, the universal fault model is established, which is classified into four groups： memory fault, CPU fault, communication fault, and service fault. A software implemented fault injection architecture （SWIFIA） for the distributed real-time systems is proposed, which is classified into three levels： target system, communication network, and software implemented fault injection system （SWIFIS）. The SWIFIS is designed with three parts： software implemented fault injector, data collection module, and fault data analysis module. The process of software implemented fault injection based on the architecture is illustrated. SWIFIS is implemented and analyzed with experiments. The major faults detected in the experiments are communication faults, memory faults, and CPU faults while the coverage rates were 37.68%, 15.47%, and 15.17%, respectively. The experimental results demonstrate that this architecture is suitable for the applications under distributed real-time environment. They offer theoretical base and evidence for further research of software dependability validation.

关 键 词：分布实时系统 系统可信验证平台 软件故障注入 故障模型 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]