Incorporation of multiple errors in reliability modeling of fault-tolerant systems

Many experiments have shown that most hardware faults do not cause immediately detectable errors. A significant proportion of the faults injected during experiments remained latent. The system operations, in which a hardware fault occurs, will remain error-free until the latent errors become effective. An effective error may, and in general does, propagate from component to another; and by propagating, an error creates other new errors. The authors develop an analytic model to study the effect of this snowball phenomenon on fault-tolerant systems. Some of the results of the sensitivity analysis are presented