An effective approach for achieving fault tolerance in hypercubes
Khalid Al-Tawil, D.R. Avresky · 2002
The hypercube network is an attractive structure for parallel processing because of its regularity. The problem of tolerating faulty processors in hypercubes has been studied by many researchers, either by using spares or by reconfiguration. In this paper, we present algorithms for achieving fault tolerance in hypercubes using spanning trees, without requiring additional spare nodes. We present two algorithms; one uses completely unbalanced spanning trees (CUST) and the other uses balanced spanning trees (BST). Both algorithms use, at most, one used link and one unused link for every reconstructed path in the reconfigured hypercube. The algorithms are optimal, in terms of the reconfiguration time and may increase the congestion of a link by, at most, one with no extra-dilation. Single-fault coverage of 100% and almost 100% fault coverage of double and triple faults are achieved by the proposed algorithms for hypercubes having a dimension of n/spl ges/10.