A novel dynamic task scheduling strategy for computational grid

Task scheduling is one of the core concerns in the computational grid environment and has been proven NP-complete. In this paper, a Dynamic Task Scheduling (DTS) strategy for computational grid environment has been proposed to minimize makespan of batch of tasks by enhancing Task Level Parallelism (TLP). The strategy works in two phases. In the first phase job is divided into tasks to exploits parallelism. In the second phase, as tasks are available in memory for execution, these tasks are allocated to the most suitable and available resources dynamically as per the strategy. Efficiency of proposed work is evaluated by designing and implementing DTS in grid simulation toolkit (Gridsim) and comparing with other state of art technique in different conditions i.e. varying the number of jobs, the number of resources and the processing capability of resources with taking remaining input parameters fixed. The simulation results revealed that DTS performs better on account of makespan for all conditions under study.