基于软错误的动态程序可靠性分析和评估  

Analyses and Estimation of Dynamic Program Reliability Based on Soft Errors

作　　者：熊磊[1] 谭庆平[1] 

机构地区：[1]国防科技大学计算机学院,长沙410073

出　　处：《小型微型计算机系统》2011年第11期2163-2168,共6页Journal of Chinese Computer Systems

基　　金：部委某项目(7130829)资助

摘　　要：基于软件实现的软错误容错方法不需要硬件开销,被认为是一种高效的软错误容错方法,而动态的实现这种方法能覆盖更多种类型的程序,因而能覆盖更多的软错误.分析硬件软错误对程序执行时代码和数据的逻辑影响,并建立了硬件软错误条件下程序运行可靠性评估模型.本文的工作为基于软件动态软错误容错算法的提出提供了理论基础,也为程序可靠性的评估提供了一种方法.我们依据体系结构层硬件对指令执行的影响将硬件构件进行分类,并分析了不同的硬件构件对程序代码和数据的逻辑影响.基于软错误对程序代码和数据的影响模型,建立了软错误条件下程序运行可靠性评估模型.最后,在实验中,对软错误条件下程序影响模型和程序运行可靠性评估模型进行了验证,实验结果证明了本文的分析和评估结果.Software-based soft error tolerance method needs not additional hardware cost. It is regarded as an efficient method to tolerate soft errors. Dynamic implementation for the method can protect more types of software, so the implementation can cover more soft errors. This paper analyzes the effects of soft errors to program execution, and we build program reliability model to estimate the reliability of program during program execution. Our analyses supply a basement to tolerance algorithms which are implemented dynamically to tolerate soft errors. They also give a method to estimate program reliability under soft errors during program execution. Specifically, we classify hardware components into several classes by their behaviors on architecture level. According to the classification of hardware components, the effects of soft error to codes and data which are in program are analyzed. Based on the effects of soft error to codes and data, dynamic program reliability model is built. Finally, we design an experiment to verify our analyses and the dynamic program reliability model. The experimental results demonstrate our analyses and validate our model.

关 键 词：动态容错 软错误 程序级 动态运行 可靠性 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]