一种基于数据流分析的故障序列生成方法

故障注入是一种测评容错机制的有效方法,它通过有意识地向目标系统引入故障来加速测试容错机制的试验过程.目前大多数故障注入实验中采用的故障序列都是通过随机抽样获取的,容易造成故障潜伏,从而导致容错机制的评测结果不准确.本文提出一种基于数据流分析的故障序列生成方法.提出<故障注入位置,故障注入时间,故障类型,故障掩码>的单粒子故障模型.在此基础上,通过数据流活跃变量分析计算得出故障注入位置属性,以及定值-引用分析确定了故障注入时间属性.并以PowerPC8548处理器和它对应的Trace模拟器为实验平台,验证了该故障序列生成方法具有90%以上的失效加速比,较好地解决了故障潜伏问题.

A Fault-list Generation Approach Based on Data Flow Analysis

Fault injection is an effective technique used to evaluate fault tolerance mechanisms.It accelerates the evaluation processing by injecting faults into the system.However,these faults are typically selected at random from the total fault space of the system.These selection has one significant limitation;that is,the no response problem,which leads inaccurate evaluation of the fault tolerance mechanisms.In this paper,we propose a fault list generation approach based on data flow analysis.Firstly,a Single Event Upset(SEU) fault model is proposed,which is composed of Fault-injection location,Fault-injection time,Fault type and Fault mask.Then Fault-injection location is calculated by live variables analysis,and Fault-injection time is calculated by reaching definitions analysis.Finally,this fault list generation approach is validated to be having an accelerated failure rate of 90% on the PowerPC8548 Processor and its Trace Simulator.