Application Research Of Multi Objective Particle Swarm Algorithm For Optimizing Task Scheduling In Data Centers

With issues like global warming and energy shortage arising,the energy saving and emission reduction of Internet data center has received unprecedented attention.With the improvement of environmental protection and the implementation of sustainable development strategy,the enormous energy consumption caused by the data center has become a common concern of industry and academia.The scale of investment datacenter is increasing at a rapid pace,the energy costs have also become the important factors which affecting corporate earnings.Therefore,under the premise of guaranteed income,it is essential for society and economic to decrease emissions through researching the reasonable task scheduling to reduce energy consumption.This paper will make reducing datacenter's energy consumption and increasing service revenue as the optimization target,which make the task scheduling problem of datacenters become a multi-objective optimization problem and get the Pareto optimal solution through Multi-objective particle swarm algorithm.The main work of this paper is as follows:1.Author compared the advantages and disadvantages of existing scheduling algorithms based on analyzing datacenters' objectives and features.On the basis of energy consumption and service revenue of datacenters,we build the model of energy consumption and the model of service revenue.2.Designing the Multi-objective particle swarm algorithm.Putting forward a new strategy in external files update methods,with the operation of mutation operator based on Genetic algorithm to make the elite solution set exchange with the useful genes,in order to improve the diversity of Pareto optimal solution set in external files;The selection of global optimal particle can influence the diversity and convergence of multi-objective particle algorithm.This essay improved the selection strategy of global optimal particle,with the introduction of crowding distance mechanism on the basis of Sigma method.At the same time,Pareto dominance relation is used to choose the optimal solution,in order to improve the convergence and the diversity of the algorithm.3.Selecting two standard test function set(ZDT and DTLZ)in Matlab,and compared the traditional NSGA-? optimization algorithm and the multi-objective particle swarm optimization algorithm,then verifying the validity of the algorithm by comparing the convergence indicators,distribution indicators and time indicator;Based on the open source emulator CloudSim,the practicality is verified through energy consumption and revenue results of CloudSim,which is on the basis of multi-objective optimized task scheduling problem in the data center.The improved multi-objective particle swarm algorithm,the classic NSGA-? and the traditional multi?objective particle swarm optimization algorithm are tested,respectively..The simulation experiment result by Matlab shows,the improved algorithm can save the runtime under guaranteeing the diversity of particle swarm and the convergence of algorithm.So,we can find the optimization solution set with more accuracy from the multi-objective optimization problem.And the simulation experiment result by CloudSim shows,the improved algorithm has a better performance in the energy consumption and revenue targets.