The buddy strategy, which is based on the buddy system, has been proposed to allocate the processors of the n-dimensional hypercube (or an n-cube) network to incoming requests. The buddy strategy can detect only 2n-k subcubes of dimension k out of a total of C(k)n2n-k subcubes. Consequently, the overall network utilization is degraded owing to the use of the buddy strategy, The Gray code strategy has been proposed as an alternative to the buddy strategy; it has double the subcube recognition ability of the buddy strategy. In this paper, we describe a new strategy that is a variation of the buddy strategy. We refer to this new strategy as the Hyper-buddy strategy. The new strategy is designed to take into account the interconnection pattern of the hypercube. It computes its buddy in a series of steps. At each step, a sub-buddy in dimension i less-than-or-equal-to n is computed. The sub-buddies are then combined to produce the final buddy, which is used to allocate the subcube. We shall show that the Hyper-buddy strategy outperforms both the buddy and the Gray code strategies. In particular, the new strategy can recognize 2n-SIGMA(p=0)k-1C(p)n. This added subcube recognition reduces system fragmentation and enhances system throughput.
