A novel task scheduling algorithm for parallel system
Zaki Ahmad Khan, Jamshed Siddiqui, Abdus Samad · International Conference on Computing for Sustainable Global Development · 2016
The parallel computing is extremely feasible as well as scalable substitutes to high performance supercomputing. In the parallel system, scheduling and mapping of task on couple of nodes is deliberated as the grave issue. The parallel system without proper task scheduling algorithm may nullify the benefits of parallelization. The present paper, proposed a novel scheduling algorithm that remedying the communication overhead, delay in task execution and in efficient processor utilization and hence, improve the total execution time. The proposed algorithm is tested and implemented on number of processor which is connected through convinced interconnection network. The performance is evaluated for linear as well as cube based parallel system with performance parameters such as Load Imbalance Factor (LIF's) as well as Execution Time. The proposed scheduling algorithm is showing better simulation results with the comparisons of other revealed scheduling algorithm on linear and cube based multiprocessor interconnection network.