A clustering model for multicast on hypercube network

Abstract

Multicast communication is one of the general patterns of collective communication in multiprocessors. On hypercube network, the optimal multicast tree problem is NP-hard and all existing multicast algorithms are heuristic. We find that the existing works are far away from optimal. Aiming to reduce the communication traffic, we propose a general clustering model for multicast on hypercube networks. Based on the model, we can construct more optimal or approximately optimal multicast algorithm for hypercube interconnection networks. And the model is also universal for existed works. © 2008 Springer-Verlag Berlin Heidelberg.