Abstract:

The need for high computation power, as well as advances in the communications technology have resulted in the rapid development of high-performance message passing architectures. Performance evaluation including scalability, fault tolerance of distributed systems, network algorithms, scheduling, load balancing and cost issues in parallel processing systems have proven to be a challenging task which is mainly considered as measures to identify the suitability of the topology for a given application.

In this research work, most common message passing architectures including the Augmented Hypercube Torus (AHT) are compared with a newly developed topology called Super-Super Hypercube 4-Cube (MS3H4-Cube) architecture. Its expandability as well as Average Routing Distance (ARD) and hardware cost are evaluated for various network sizes and component cost ratios. OPNET package will be utilised to perform the modeling and simulation of the system architecture and task allocation schemes. The author intends to use the Artificial Intelligence applicable to the newly developed task allocation scheme in order to further improve its applicability and flexibility. This would follow by a benchmarking to validate the results claimed in this research work.