A Methodology to Calculate a Program's Robustness against Transient Faults

摘要

Computer chips implementation technologies are evolving to obtain more performance. The side effect of such a scenario is that processors are less robust than ever against transient faults. As on-chip solutions are expensive or tend to degrade processor performance, the efforts to deal with these transient faults in higher levels are increasing. Software based fault tolerance approaches against transient faults often use fault injection experiments to evaluate the behavior of applications with and without their fault detection or fault tolerance proposals. Those fault injection experiments consumes lots of CPU time by running or simulating the application being evaluated as many times as necessary to obtain a reasonable valid statistical approximation. This paper proposes the concept of a program's robustness against transient faults and presents a methodology for exhaustively calculate this robustness based on program's execution trace over an architecture and on information about the used architecture. The presented approach, besides calculating a precise robustness, accomplishes its work faster than using fault injection experiments with ±2% of confidence interval.