Evaluation methods for inter-connection structure used in parallel processing systems

The demand for high computational capacity always surpasses the available computing power that could be realized from the current technology. Available technology limits the maximum power that can be derived from a single processor. The other alternative is to go in for the time tested devicepara11eliEtn that employs a number of cooperating processors interconnected through one form or other of the many existing interconnecting networks which vary between the extremes of the single bus and the completely connected scheme. Although the para-leI computing system can conceptually provide a linear speed up over uniprocesso: system, practical systems fall short of this. One of the main reasons for this is the overhead associated with the data management and communication. In this lecture, we review the major class oxA3; interconnection networks that have been available in literature with a specific view of analyzing its architecture with respect to various metrics that contribute to the performance of parallel processing system. Finally,the architectural design and specification of a flexible cable Bus (FLEXIBUS)which is design pursued as one of the interconnecting structure for bus-based parallel processing systen and its expected performance are also discussed