Research On Energy-Efficient Scheduling Strategy For Multiple DAGs Workflow In Heterogeneous Cloud Platform

With the rapid development of cloud computing technology,the scale of cloud data center servers is expanding,and high energy consumption has become a prominent problem of cloud platform.In previous studies,the energy consumption is reduced by the migration and aggregation of virtual machines(VMs).However,when the VMs are deployed to the physical machines(PMs),it will not be migrated for a short time.At this time,it can be considered to further reduce the energy consumption of the cloud platform through task scheduling.However,most of the current cloud platform task scheduling methods aim to reduce the task execution time.When the heterogeneous cloud platform is composed of heterogeneous nodes with different computing performance,this task scheduling strategy may lead to higher energy consumption.In addition,most computing tasks on cloud platforms are represented in the form of scientific workflow,that is,the sub-tasks are not independent of each other,and the Directed Acyclic Graph(DAG)can be used to represent the dependency between tasks.In the actual cloud task scheduling,multiple workflows are often scheduled at the same time,so it is necessary to study the scheduling method for multiple DAGs workflow.Based on the above analysis,combined with the heterogeneity of the cloud platform,aiming at the scheduling of multiple DAGs workflow in the heterogeneous cloud platform,this thesis analyzes the dependence of DAG tasks,considers the relevant factors that affect the optimization of energy consumption of the heterogeneous cloud platform in task scheduling,and proposes an energy-efficient scheduling strategy of multiple DAGs workflow in the heterogeneous cloud platform with the goal of minimizing the energy consumption of the heterogeneous cloud platform.The main work of this thesis is as follows:(1)The research status of task scheduling in cloud platform are investigated,expounds the existing task scheduling methods from the perspective of optimization objectives and task types respectively,excavates the parts to be improved in the current DAG scientific workflow scheduling.Based on the summarization of the existing literature,existing problems in DAG workflow scheduling are extracted,and the energy-efficient scheduling technology for multiple DAGs workflow in heterogeneous cloud platform is analyzed.(2)The thesis constructs the DAG workflow model and task scheduling model.In the DAG workflow model,the DAG workflow is formally defined,the relevant conditions of DAG are constrained,and four kinds of DAG workflows with different structures used in the study are explained.In the task scheduling model,the related variables involved and the relationship between them is characterized.The model perceives the workload of the PM in real time according to the VM state,and predicts the busy and idle energy consumption of the PM,and then obtains the energy consumption of the cloud platform.Then we formulate the task scheduling optimization problem,which tries to minimize the energy consumption of heterogeneous cloud platform.(3)The thesis proposes a multiple DAGs workflow scheduling algorithm to minimize energy consumption of heterogeneous cloud platform.The algorithm is achieves the optimization objective through two stages in the process of workflow scheduling.Phase 1:calculate the priority of the workflow to be assigned;Phase 2: schedule the workflow to be assigned to the appropriate computing nodes according to the priority.In the first phase,we consider the combination of coarse-grained sorting with DAG workflow as scheduling unit and fine-grained sorting with sub-tasks in DAG workflow as scheduling unit to achieve the optimal sorting of workflows to be assigned,which is helpful to the implementation of task allocation in the second phase.In the second phase,a variety of factors that affect the energy consumption of heterogeneous cloud platform are comprehensively analyzed to solve the proposed optimization problem.Finally,the simulation experiment was carried out on CloudSim.The algorithm proposed in this thesis is compared with other task scheduling algorithms,and the energy-efficient strategy proposed in this thesis is added to other algorithms to improve it.The experimental results show that the proposed algorithm has a good performance in reducing the energy consumption of the cloud platform,and the improved comparison algorithm has a significantly lower energy consumption than the original algorithm.This shows that this algorithm not only minimizes the energy consumption of heterogeneous cloud platform,but also helps other task scheduling methods to further optimize energy consumption.