A limited-global-information-based multicasting scheme for faulty hypercubes

We study the multicast problem in hypercubes with faulty nodes. We assume that each node has limited global information of fault distribution, which is captured by the safety level associated with the node. Basically, the safety level is an approximate measure of the distribution of faulty nodes in the neighborhood. We first propose a safety level-based multicasting scheme which guarantees time optimality. We then present a variant of this scheme that reduces the number of traffic steps. Finally, an address-sum-based multicasting scheme is studied which utilizes both the distribution of destination nodes and the safety level of neighboring nodes. Simulation results show that the traffic generated in both schemes is very close to optimal. Time optimality is guaranteed when the source is safe and at most two additional time steps are required when the source is unsafe in an n-dimensional cube (denoted as n-cube) with up to n-1 node faults