FAULT-TOLERANT GOSSIPING ON HYPERCUBE MULTICOMPUTERS

Various algorithms for reliable gossiping in faulty n dimensional hypercube multicomputers are described and analyzed. The goal is that each processor receives complete information from all the other processors even in the presence of faults. One of the main characteristic of the algorithms is that no information on the identity of the faulty nodels/links is required. The exchange between any two processors is realized such that the data moves through disjoint paths. We propose solutions designed for systems which use store and forward models of communication. The proposed algorithms are parametrized by the maximum number of faults that they can take into account. In all cases, they approach the minimum time complexity.