Self-stabilizing wormhole routing in hypercubes

Wormhole routing is an efficient technique used to communicate message packets between processors when they are not completely connected. To the best of our knowledge, this is the first attempt at designing a self-stabilizing wormhole routing algorithm for hypercubes. Our first algorithm handles all types of faults except for node/link failures. This algorithm achieves optimality in terms of routing path length by following only the preferred dimensions. In an n-dimensional hypercube, those dimensions in which source and destination address bits differ are called preferred dimensions. Our second algorithm handles topological changes. We propose an efficient scheme of rerouting flits in case of node/link failures. Similar to the first algorithm, this algorithm also tries to follow preferred dimensions if they are nonfaulty at the time of transmitting the flits. However, due to topological faults it is necessary to take non-preferred dimensions resulting in suboptimality of path selection. Formal proof of correctness for both solutions is given. (Abstract shortened by UMI.)