Title :
Probability, graphs, electrical networks and computer fault diagnosis
Author_Institution :
Dept. of Electr. Eng., New Brunswick Univ., Fredericton, NB, Canada
Abstract :
A method for the identification of faulty interprocessor links in an arbitrarily large multiprocessor assembly is described. The description given and the associated examples refer to the single fault case, but the procedure is readily adapted to multiple faults and to processor (vertex) fault diagnosis. The diagnostic process is probabilistic and compares a sequence of faulty computational substructures which may be, as used here, computational trees in the graph of the system. The various probabilities used in the analysis are obtainable numerically as electrical resistance values in a resistor network having the same structure as the system graph
Keywords :
multiprocessor interconnection networks; performance evaluation; probability; system recovery; trees (mathematics); computational trees; computer fault diagnosis; faulty computational substructures; interprocessor links; multiple faults; multiprocessor assembly; probabilities; resistor network; single fault case; vertex fault diagnosis; Computer architecture; Computer networks; Electric resistance; Fault diagnosis; Hypercubes; Large-scale systems; Multiprocessing systems; Probability; Resistors; Tree graphs;
Conference_Titel :
Circuits and Systems, 1992. ISCAS '92. Proceedings., 1992 IEEE International Symposium on
Conference_Location :
San Diego, CA
Print_ISBN :
0-7803-0593-0
DOI :
10.1109/ISCAS.1992.229928