A Cost-Performance Scalability Measure for Interconnection Networks and a Novel Scalable Cube-Based Topology.

Employing proper interconnection networks is essential for enhancing the overall performance of multiprocessor systems. Efficient size scalability is a critical consideration when designing interconnection networks for cost-effective implementation and management of communication overhead. Size scalability indicates how well systems can scale capacity when expanding to permitted larger configurations based on the network topology constraints. In this paper, we first explore establishing a formulation for evaluating cost-performance scalability. Then, a comparison among hypercube, as a prominent and widely used interconnection network topology, and its famous varieties is conducted given their scalability. We then propose a novel variation of the hypercube, named Overlapped Cube (or OCube, for short) on m , n , k ∈ ℕ such that m ≥ n > k and denoted by O Q m , n , k and examine its topological properties. Comparing the scalability of some traditional networks and known variants of hypercubes demonstrates that OCube has the potential to be an effective interconnection network with desirable scaling behavior. [ABSTRACT FROM AUTHOR]