Total System Cost Analysis and Comparison of DX-Tree and Star Ring Architectures in conjunction with Master-Slave Multi-Super-Hypercube topology

Abstract

In todays advanced technology environment, we come across a number of research areas that require extensive data processing. These areas include but not limited to, modeling and simulation of accuracy, safety and reliability of nuclear weapons, global warming, weather forecasting, DNA computing, nanotechnology, immune cell system, optical computing. Since we constantly continue to explore new avenues in our research program, hence the demand for improvements in other scientific and engineering parameters are continuously on the rise. These scientific and engineering requirements include but not limited to speed, reliability, fault tolerance, availability, compatibility, scalability, flexibility, cost to name a few. In todays environment, utilisation of parallel processing not only satisfies increasing demand for extensive data processing, yet it relatively fulfils deficiencies that one would experience in these scientific and engineering requirements. To this effect the authors have introduced architectures based on Master-Slave Multi-Super-Hypercube DX-Tree architecture and Master-Slave Multi-SuperHypercube Star Ring Architecture. For these architectures and the current message passing architectures, the total system costs are developed and through mathematical modeling and simulation are compared. In this comparison the merit and demerits of these architectures from the cost point of view are highlighted to enable the user to select an appropriate Message-Passing architecture which would best satisfy ones scientific requirements.