DVFS ENABLED GREEDY MECHANISM FOR ENERGY AWARE TASK SCHEDULING IN HIGH PERFORMANCE COMPUTING SYSTEMS

Nowadays, the high performance computing (HPC) systems are widely adopted in data centers to process and accomplish the tasks which consume more resources. These tasks increase load on the data centers however, HPC systems utilize multicore processors which facilitate faster computing. However, inappropriate use of these multicore systems leads to excessive energy consumption and wastage of computing resource. Dynamic voltage and frequency scaling is one of the promising solution to overcome energy consumption related issue in most of the processor based systems. In this paper, we present a combined energy aware scheduling model to schedule multiple types of heterogeneous tasks along with DVFS for high performance computing platform. We introduce a schedule architecture model which acts as a central scheduler along with DVFS. First of all, the incoming tasks are analyzed based on their deadline constraints and resource requirement. DVFS scheme is applied to adjust the voltage and frequency based on the core’s configuration. Later, the proposed greedy energy aware scheduling approach assigns the tasks to the specific processor, based on the response time and probability value of the task. This problem is formulated as an optimization process. Moreover, the proposed technique minimizes the queuing delay, which helps to minimize the delay in scheduling. The comparative analysis shows that, the average waiting time and average energy saving ratio of the proposed approach is improved