Research And Implementation Of Energy Saving Method Based On Resource And Task In Cloud Platform

With the rapid development and the widespread application of cloud computing,the huge power consumption of server cluster becomes the focus of people's attention gradually.The greater the requirement of cloud services,the greater the demand for the size of infrastructure cluster,and the energy consumption problem is more serious.Aiming at this kind of high energy consumption problem,this paper puts forward an energy saving method based on combination of resources and task.This method reduces the energy consumption of the cloud center effectively.This paper analyses the energy bottleneck in cloud computing environment firstly,then compares the two kinds of resource consumption forecasting techniques,and compares a variety of task allocation strategies,next summarizes the advantages,disadvantages,and application scenarios of all kinds of energy saving methods,finally identifies a new way of thinking to solve the problem of energy consumption:first of all,forecast the resource consumption of the tasks in task queue,then implement the task allocation strategy based on forecast data to use resources reasonably,and reduce energy consumption of cluster.This article first has been clear about the various factors affecting task resource consumption,and then proposes a resource prediction model based on task granularity.The model consists of three parts:feature extraction and data acquisition,data purification,neural network prediction.Data purification can effectively reduce the noise and improve the prediction accuracy,and the improved neural networks can improve the prediction accuracy very well.Task allocation strategy based on resource prediction mainly involves two key problems:virtual machine selection and virtual machine migration.This paper puts forward two virtual machine selection algorithms:Max-Match algorithm and Min-Cost algorithm,and two virtual machine migration algorithms based on Best Fit Decreasing(BFD)adaptive algorithm:Min-Power BFD adaptive algorithm and Threshold-based BFD adaptive algorithm.Task allocation strategy consists of the two kinds of algorithms to implement reasonable task scheduling.Experiments contains two parts.The first part proves the accuracy of the resource prediction model,and evaluates the convergence and performance of every algorithm in model.Results show that the model has good adaptability and high accuracy.The second part verifies the energy saving effect of task allocation strategy,and contrasts the applicable scenarios of various strategies.In this paper,the energy saving method based on resources prediction and task allocation not only has good effect on consumption,but also provides a new way of thinking about related research in the future.