A Proposed Model: Linearly Extensible Triplet Network (LETN)

Abstract

Past few years, there has been keen interest over computation power and scalability of the computer system. The computation power can be increased massively in two ways either in terms of optimizing hardware by increasing the number of processor (nodes) or in terms of optimizing software by structuring the program in such a way that the large task is subdivided into independent subtasks (load). The scheduling and mapping of tasks on the set of processors has significant role in parallel and distributed computing system. The scheduling problem involves assigning a balanced task to each processor according to its performance capabilities. It also minimizes communication overhead in the network. To speed up computation power, a linearly scalable triplet based multiprocessor architecture (network) has been proposed here. It has been designed to meet the current trends in achieving high computation power in multiprocessor architecture. It allows the extension of the network by adding new subsystems of three processors at a time which increases the parallel computation. The properties and task Allocation algorithm of proposed model has been discussed deeply. By discussion and comparison authors come to the conclusion that proposed model is better and is need of the time.