计算机工程
計算機工程
계산궤공정
COMPUTER ENGINEERING
2015年
2期
57-62,75
,共7页
胡倩%王超%王海霞%汪东升
鬍倩%王超%王海霞%汪東升
호천%왕초%왕해하%왕동승
故障注入%系统可靠性%故障模型%故障检测%结构级模拟器
故障註入%繫統可靠性%故障模型%故障檢測%結構級模擬器
고장주입%계통가고성%고장모형%고장검측%결구급모의기
fault injection%system reliability%fault model%fault detection%structure level simulator
故障注入技术是评价系统可靠性的有效方法。现有基于仿真的故障注入平台大多基于现场可编程门阵列或超高速集成电路硬件描述语言实现,对故障模型的支持非常有限。为此,基于Simics结构级模拟器,设计并实现系统级硬件故障注入平台。该平台上层支持不同固件、操作系统以及应用程序,底层支持对处理器典型流水部件的故障注入,同时实现瞬时故障、永久故障和间歇故障模型以及其他较全面的故障类型,并将一组系统级故障检测机制集成入平台中。实验通过监测硬件故障在系统级的传播,对比分析了故障对不同部件造成的系统级影响,结果表明,瞬时故障对系统影响较小,永久故障容易引起系统失效,间歇故障对各部件有不同程度的干扰作用。
故障註入技術是評價繫統可靠性的有效方法。現有基于倣真的故障註入平檯大多基于現場可編程門陣列或超高速集成電路硬件描述語言實現,對故障模型的支持非常有限。為此,基于Simics結構級模擬器,設計併實現繫統級硬件故障註入平檯。該平檯上層支持不同固件、操作繫統以及應用程序,底層支持對處理器典型流水部件的故障註入,同時實現瞬時故障、永久故障和間歇故障模型以及其他較全麵的故障類型,併將一組繫統級故障檢測機製集成入平檯中。實驗通過鑑測硬件故障在繫統級的傳播,對比分析瞭故障對不同部件造成的繫統級影響,結果錶明,瞬時故障對繫統影響較小,永久故障容易引起繫統失效,間歇故障對各部件有不同程度的榦擾作用。
고장주입기술시평개계통가고성적유효방법。현유기우방진적고장주입평태대다기우현장가편정문진렬혹초고속집성전로경건묘술어언실현,대고장모형적지지비상유한。위차,기우Simics결구급모의기,설계병실현계통급경건고장주입평태。해평태상층지지불동고건、조작계통이급응용정서,저층지지대처리기전형류수부건적고장주입,동시실현순시고장、영구고장화간헐고장모형이급기타교전면적고장류형,병장일조계통급고장검측궤제집성입평태중。실험통과감측경건고장재계통급적전파,대비분석료고장대불동부건조성적계통급영향,결과표명,순시고장대계통영향교소,영구고장용역인기계통실효,간헐고장대각부건유불동정도적간우작용。
Fault injection provides an effective method to evaluate the reliability of system,which is a complex topic in multicore situation. There are many simulation-based fault injection tools now,most of which are implemented by Field Programmable Gate Array(FPGA) and Very High Speed Integrated Circuits Hardware Description Language(VHDL), with limited fault models. Based on the widely used system simulator Simics in computer architecture,this paper designs and implements a system level fault injection platform,supporting different firmware,OS and applications. It can inject faults into several components,with different fault models( including transient faults,permanent and intermittent faults) and most fault types. Further more,it integrates fault detection module into the system. After observing of the propagation of hardware faults in system,it analyzes the effect of different components,fault models on system level,inspiring fault detection,and finds that transient faults have a little impact on system, while permanent faults seriously interrupt the running and intermittent faults performs differently on different components.