A model for the analysis of the fault injection process

Abstract

Results of fault injection experiments performed under different conditions can only be related to each other, if their interpretation is based on a thorough understanding of activation and propagation of faults and errors. We analyze these processes by applying a special layer model of a computing system. Our aim is to model the transformation of a fault on a signal line into a system failure as the propagation of erroneous information through multiple layers. Two specific layers that describe the fault activation process have been sufficiently completed and are presented here. A quantification for these is derived and different applications are summarized. Excellent correspondence between analytical results based on modeling and experimental data is found. A prediction of fault activation with high accuracy is possible, as well as a quantitative evaluation of the effect of synchronizing fault injection.<>