EXTENDED BANYAN HYPERCUBE

Abstract

Banyan hypercube network is a static network, In which various hypercube are interconnected In banyan fashion. Banyan hypercube network has good embedding and partitioning ability, besides lowering the cost factor and average distance as compared to other static interconnection networks. However, it has only limited number of processors in any specific configuration. When new processors are added in a. saturated banyan hypercube network, the network configuration has to be changed and whole hardware is to be rebuild accordingly. This puts a limit on the use of banyan hypercube network where extendability is desired. In this thesis Extended banyan hypercube, a new network is proposed to overcome the extendability problem of banyan hypecube network. Any number of new processor nodes can be added to any given configuration ofextended banyan hypercube with very low extendability cost. Extended banyan hypercube is compared with banyan hypercube in terms of properties and embedding. A generalised mapping strategy is being used to embed common problem graphs. The results show that Extended banyan hypercube has lower degree and diameter and hence lower cost factor when compared to banyan hypercube network.- They show equal embedding properties as that of banyan hypercube network and can embed trees, meshes, rings, pyramids and multiple pyramids (as problem graphs), efficiently.