FAULT INJECTION EXPERIMENTS USING FIAT

This paper presents the results of several experiments conducted using the FIAT (fault injection-based automated testing) system. FIAT is capable of emulating a variety of distributed system architectures, and provides the capabilities to monitor system behavior and inject faults for the purpose of experimental characterization and validation of a system's dependability. The experiments consisted of exhaustively injecting three separate fault types into various locations, encompassing both the code and data portions of memory images, of two distinct applications executed with several different data values and sizes. Fault types were variations of memory bit faults. The results have shown that there are a limited number of system-level fault manifestations. These manifestations follow a normal distribution for each fault type. Error detection latencies were also found to be normally distributed. Thus, sampling techniques can be used to cut down the size of the fault injection space without sacrificing fidelity. Finally, the methodology can be used to predict the system-level fault responses during the system design stage. As an example, a model is developed for the occurrence of illegal opcode traps, based on the instruction set architecture, a static analysis of the workload, and a model of the low-level faults. The prediction of system-level responses is shown to be within 10 percent of the experimental data. © 1990 IEEE