The Research Of An Efficient Energy Scheduling Algorithm For Workflow Tasks In DVFS-enabled Heterogeneous Environment

In recent few years,with its super-efficient computing power and data storage capacity,the cloud computing technology is in full swing,and the explosive growth of energy consumption in cloud centers has become a critical issue,for going against the advocated green computing.How to make a better trade-off value between the performance and reducing energy dissipation has become a major challenge for both academia and industry.Besides improving the hardware device to reduce the energy consumption,the research of workflow task scheduling algorithm for saving energy while still ensuring timeliness has also been concerned widely.The main objective of the energy saving optimization mechanisms of the existing parallel task scheduling is to reduce the number of the resource used and the time occupied and improve the utilization of resource,so as to achieve a balance between system performance and power consumption.It is based on making all tasks in accordance with the topology of their dependencies into groups,and then assigns each group of tasks to their appropriate time slots on the processors.Moreover,various mechanisms for reducing the energy consumption have been investigated in the earlier work.Such as Dynamic Voltage/Frequency Scaling(DVFS)and Dynamic Power Management(DPM).DVFS technique enables processors scale down their voltages and clock frequencies during the idle periods,and DPM turns the idle components off leading the resource to the hibernate mode to reduce the power consumption while the idle time is long enough.With still meeting the performance-based service level agreement,this paper proposes a new energy aware scheduling algorithm named DVFS-enabled Efficient energy Workflow Task Scheduling(DEWTS)striving to optimize simultaneously the energy savings through DVFS technique for parallel applications in the heterogeneous distributed computing systems.This method firstly makes good use of the slack time by reducing some potential processors that execute tasks inefficiently to save the total energy consumption,and secondly,schedules the tasks on the DVFS-enabled processors to reduce processor energy whenever there have residual slack time after processors merging phase.To evaluate the performance of our proposed approach,we choose the CloudSim simulator as our experiment environment and simulate four groups of DVFS-enabled heterogeneous processors for our study.Based on amount of randomly generated DAGs workflows,the experimental results show that DEWTS can reduce the total power consumption with various parallel applications as well as balancing the scheduling performance.