Research On Energy Consumption Optimization Task Scheduling Method For CPU-GPU Heterogeneous Cloud Platform

With the rapid development of the cloud computing market,data centers are also ushering in an explosive period,accompanied by the continuous increase of carbon emissions.In order to optimize the industrial structure and energy structure,the idea of energy saving and emission reduction has become a key factor for cloud computing in the digital transformation of enterprises.In cloud computing,task scheduling is one of its main technologies.Reasonable task scheduling technology can effectively reduce the power consumption of cloud platform,while traditional task scheduling algorithms pay less attention to the collaborative work of CPU and GPU,which leads to the computing resources of the system.cannot be fully utilized.Based on the above analysis,this paper combines the heterogeneity of cloud platforms,schedules multiple task workflows on heterogeneous cloud platforms,analyzes the synergy between CPU and GPU,and considers the resource utilization and execution time of CPU and GPU.Minimizing the energy consumption of heterogeneous cloud platforms is the research goal,and a task scheduling strategy for energy consumption optimization of CPU-GPU heterogeneous cloud platforms is proposed.The main work done is as follows:(1)The research status of task scheduling methods for cloud platforms at home and abroad is studied.Firstly,it briefly analyzes the concept of task scheduling algorithm of cloud platform at home and abroad and the reasons for the generation of heterogeneous cloud,and then expounds its social value.This paper starts with several energy-saving and emissionreduction technologies of cloud computing,systematically summarizes the research status of task scheduling algorithms,and points out the shortcomings of current research.Finally,the areas for improvement are identified and the objectives to be achieved by this study are proposed.(2)Modeling of task scheduling model and energy consumption model in heterogeneous cloud platforms.Thinking about the heterogeneity in the cloud platform environment,it is found that reasonable task allocation is the key issue to improve the execution efficiency of the heterogeneous cloud platform;in the task scheduling model,the deadline of subtasks is used as a constraint,and real-time according to the status of task scheduling Perceive the resource utilization of physical machines,and dynamically adjust the execution order of subtasks to complete the optimization of time;build a task scheduling model with the execution time of subtasks as the research goal,and find out the difference between task scheduling energy consumption and task scheduling under heterogeneous cloud platforms.The relationship between the CPU-GPU resource utilization and the energy consumption model under the heterogeneous cloud platform are established.(3)Research on multi-task scheduling strategy in heterogeneous cloud platform.Firstly,the importance of the multi-task scheduling algorithm in the heterogeneous cloud platform is expounded,and the definition of the multi-task scheduling strategy is briefly summarized,and the multi-task scheduling algorithm in this paper is proposed.To improve the optimization goal,the first aspect: reasonable division of multi-task data flow;the second aspect: dynamic load balancing adjustment of multi-task data flow,through two methods to improve the execution efficiency of the pipeline and reduce the cloud platform.The purpose of task scheduling completion time and reducing overall energy consumption.Finally,simulation experiments were carried out on cloud simulation software.The HCGTS scheduling algorithm proposed in this paper,the Min-Min scheduling algorithm,the MaxMin task scheduling algorithm,and the PSO algorithm are set up in a comparative experiment under multiple indicators to verify the task scheduling algorithm proposed in this paper.Superiority in energy conservation and emission reduction,resource utilization,and execution efficiency.