Fault-tolerance on Boolean n-cube architectures

Abstract

An approach to fault-tolerant Boolean n-cube architectures (FTBns) is proposed in this paper. We employ spares, including nodes, links and switches, to reconfigure a failed system so that system topology with its original dimension can be retained. The FTBn is designed in two levels. In the first level, we use a Boolean m-cube of 2m nodes with 2P, p ≤ m, spare nodes, and some switching elements to build a fault-tolerant module (FFM). Then an FTBn, n ≥ m, is built in the second level by taking 2n-m FTMs, and augmenting several switching elements between two adjacent FTMs. We will show that each FTM can achieve full spare utilization, and also that the degree of each node maintains a constant n. A two-phase reconfiguration algorithm is developed to allocate an adequate spare node to replace a faulty node. Finally, the reliability and costs of the FTBn are evaluated, and we then show that the FTBn can achieve higher or the same reliability as previous comparable systems at less extra hardware cost.