D. Blough, G. Sullivan, and G. Masson. Fault diagnosis for sparsely interconnected multiprocessor systems. In 19th Int. IEEE Symp. on FaultTolerant Computing, pages 62--69. IEEE Computer Society, 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....can occur in a massively parallel system: 1) the faults are scattered throughout the system, separated from each other, and (2) the faults are located close to each other forming a group. In most practical cases both situations can be handled using just a portion of the diagnostic information [3]. The other performance bottleneck originates in the classification of those units which are not involved in a contradiction and whose fault state cannot be surely identified. Deterministic algorithms require complex methods for this task, since they must guarantee a correct and complete diagnosis ....

D. Blough, G. Sullivan, and G. Masson. Fault diagnosis for sparsely interconnected multiprocessor systems. In 19th Int. IEEE Symp. on FaultTolerant Computing, pages 62--69. IEEE Computer Society, 1989.

....limit, the algorithm classifies unit u i as fault free. To provide asymptotically correct diagnosis, the algorithm conducts multiple tests on the same unit. On the other hand, the Majority algorithm by Blough et al. considers the k i sum as the result of a majority vote for faulty classification [7]. The paper shows, that using this simple diagnosis strategy correct diagnosis is assured with high probability in a class of systems that includes hypercube architecture. Another important issue is the space and computational complexity of diagnosis. The diagnosis problem in its most general ....

D. Blough, G. Sullivan, and G. Masson, "Fault Diagnosis for Sparsely Interconnected Multiprocessor Systems," 19th Int. IEEE Symp. on Fault-Tolerant Computing, pp. 62-69, 1989.

....mesh systems. Furthermore, some specific fault patterns such as scattered faults or entire rows or columns being faulty can be easily diagnosed with these algorithms. Algorithms based on this strategy are applicable to most regular interconnected networks. Recently, Blough, Sullivan and Masson [3] have also proposed a diagnosis algorithm where each unit forms a private opinion of its neighboring units, based on its interaction with them. A diagnosis algorithm that consists of simple majority voting among the neighbors of a unit to determine the status of that unit is then executed. The ....

....step in this algorithm is similar to the confidence level generation step in the algorithm previously. However, the second step in this algorithm yields inferior results in comparison to those in [39] The performance of a system with a majority voting diagnosis algorithm has also been analyzed in [3]. It has been shown that such a diagnosis algorithm can identify faulty units in the system with a very high probability if each unit is connected at least to O(logjU j) other units.z 6 Summary and Conclusions The area of system level diagnosis has experienced a great development in the last two ....

D.M. Blough, G.F. Sullivan and G.M. Masson, "Fault diagnosis for sparsely interconnected multiprocessor systems," IEEE Symp. Fault-Tolerant Comput., 1989, pp. 62-69.

....in [18] This system model avoids many of the pitfalls of the PMC model, including the need for the complete tests, the permanent nature of faults, off line testing, and an upper bound on the number of simultaneously faulty subunits. More recent work on probabilistic diagnosis can be found in [5, 11, 12, 30, 36]. 3 Comparison The goal of this paper is to compare membership algorithms with system diagnosis algorithms. Since membership algorithms typically are distributed by nature, we concentrate the comparison on distributed system diagnosis. In principle, membership and system diagnosis deal with a ....

D. Blough, G. Sullivan, and G. Masson. Fault diagnosis for sparsely interconnected multiprocessor systems. In Proceedingsof the 19th International Symposium on FaultTolerant Computing, pages 62 -- 69, Jun 1989.

....This system model avoids many of the pitfalls of the PMC model, including the need for the complete tests, the permanent nature of faults, off line testing, and an upper bound on the number of simultaneously faulty subunits. More recent work on probabilistic diagnosis can be found in [BP90a, BSM89, BP90b, LYS93, Pel93] 6.2 Comparison We argue that the major differences between algorithms traditionally viewed as system diagnosis or membership are in the failure model and the strength of the properties provided by the service. Since membership algorithms typically are distributed by ....

D. Blough, G. Sullivan, and G. Masson. Fault diagnosis for sparsely interconnected multiprocessor systems. In Proceedings of the 19th Symposium on Fault-Tolerant Computing, pages 62 -- 69, Jun 1989.

....Denning s intrusion detection model is NIDX [3] The IDES model is very system independent. NIDX extends this model to include system dependent knowledge such as a description of file systems and rules regarding system policies. 1.5. 6 Network Management Expert Systems Blough, Sullivan, and Masson [6] describe an algorithm for multiprocessor fault diagnosis that requires only a limited number of testing interconnections between processors in the network. Each unit in the network forms a private opinion of the status of its neighboring units. The diagnosis algorithm consists of taking a ....

Blough, D. M., Sullivan, G. F., and Masson, G. M. Fault diagnosis for sparsely interconnected multiprocessor systems. In Proceedings of the Nineteenth International Symposium on Fault-Tolerant Computing (1989), IEEE Computer Society, IEEE Computer Society Press, pp. 62--69.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC