Fault Node Identification Using Comparison Models

To detect fault nodes, multiprocessor systems requires rapid and literal mechanisms. The main problem of system level fault diagnosis is difficult and not efficient and also no such generic deterministic solutions are known for motivating the purpose of heuristic algorithms. In this paper we are showing how artificial unaffected systems (AUS) can also be used for fault diagnosis in multiprocessor systems having large number of nodes. Here we deal with two models, (i) The generalized comparison model, (ii) The simple comparison model, and also we propose AUS based algorithms for identifying some faults in diagnosable systems and based variation among units. We conducted experimental analysis of these algorithms by reproduce them on randomly generated diagnosable systems of different sizes under different fault scenarios. These results indicate that the AUS based approach provides a better solution to the system level fault diagnosis