
Consider an NMR system with N-Version software, and a non-diagnosing medianselect voter. There is a non-zero "correlation" probability that a fault in one software module also occurs in another software module simultaneously. The probability of a triple software fault is nil. The system has the following parameters: λ2= Processor fault rate, λs= Software module fault rate, λy= Voter fault rate, Pc= Probability that a software fault affects two versions. Draw a SAN model (Petrinet) for determining the unreliability of this system. Be sure all timed-transitions, transition rates, and inhibitor arcs (input gates) are clearly shown and defined. Notes: - To draw the model, you may need to take advantage of the following submodel (example). The model shows that once the transition T1 fires, e.g., due to a component failure, a token will be placed in P2 or P3 with probability of 0.7 or 0.3 , respectively. For instance, a token will be placed in P2 with probability of 0.7 and no token in P3. - A median-select, after sorting, simply selects the middle value from N results (Even if a processor crashes completely, its null result will be one of the N results). The median-select ensures the middle value selected is within the range of the correct results - Obviously this happens if at least half of the versions are still functioning. - A non-diagnosing voter means that the voter does not attempt to diagnose which version result is correct, or whether the median selected is correct or not.