Scalable architectures with k-ary n-cube cluster-c organization

Recent advancements in VLSI and packaging technologies demonstrate attractiveness in building scalable parallel systems using clustered configurations while exploiting communication locality. Clustered architectures using buses or MINs as the inter-cluster interconnection do not satisfy both the above objectives. This paper proposes a new class of k-ary n-cube cluster-c scalable architectures by combining the scalability of k-ary n-cube wormhole-routed networks with the cost-effectiveness of processor cluster designs. This paper focuses on direct cluster interconnection. The interplay between various system parameters and routing schemes are analyzed to determine optimal configurations under the constant bisection bandwidth constraint. Our analysis indicates that small sized clusters with a ring intra-cluster topology and a 2D/3D/4D inter-cluster network connecting these clusters offer best system performance.<
>