A VARIANCE-REDUCTION TECHNIQUE VIA FAULT-EXPANSION FOR FAULT-COVERAGEESTIMATION

The estimation of fault coverage (FC) far ultra dependable systems is a daunting task. Typically, system FC is estimated via experimental te chniques such as fault injection, and the gathered data are analyzed u sing statistical models, Specifically, faults are randomly selected, t hen injected Ei into the system, and the response of the system is rec orded. If the injected fault is detected, then the result is recorded as a 1; otherwise it is a 0. A point estimate and s-confidence interva l are then derived from the experimental data, The difficulty with thi s approach is that ultra-dependable systems have FC, C greater than or equal to 1 - 10(-5). To estimate C accurately requires more than 10(i ) data points, i = -log(10)(1-C). A technique for enumerating equivale nt fault classes can be used to reduce the number of required experime nts. The enumeration process is fault expansion, which determines the set of equivalent faults via an analysis of the system structure. This paper presents a fault expansion (FE), variance reduction technique ( VRT) that uses the expanded fault data to calculate a point estimate a nd confidence interval for the fault detection coverage. This FE-VRT c an reduce appreciably the number of fault injection lion experiments r equired to estimate C for an ultra-dependable system. Typically, perfo rming fault injection experiments is costly, in terms of both process time and computer resources. Fault injection results and the equivalen t expanded fault-set for each fault are included in this paper to demo nstrate the power of FE-VRT. FE-VRT is a viable method for increasing the accuracy of a FC estimate.