A Unified Model for Sub-System Allocation on Product Networks

Product networks are a class of interconnection networks that includes many useful networks such as multi-dimensional mesh, multi-dimensional torus, hypercube, and generalize hypercube. Most existing processor allocation methods are proposed for a specific interconnection network such as 2-D mesh or hypercube. In this research study, a “unified model” for processor allocation is proposed for partitionable product networks such as multi-dimensional mesh, multidimensional torus by using “a k-tree system state representation”. This unified model includes a general methodology of network partitioning, combining, searching, and best fit criteria, which can be scalable for any network belonging to the product networks class. It will be show that this model can be efficiently applied to existing useful product networks such as 2-D and 3-D meshes and toruses, yielding smaller time complexity and higher system performance than existing meshbased algorithms.