Research On Data Center Energy Optimization And Task Scheduling Methods Based On Smart Grid

With the rapid development of Internet technologies such as big data and cloud computing,users are generating increasingly amount of business data,forcing Internet service providers to store,process and maintain them on more and more data centers.Research shows that energy-saving task scheduling strategies can effectively reduce asset,operation and maintenance costs,and meanwhile data center companies are making themselves more environment-friendly by utilizing renewable energy.However,there is still seldom research that comprehensively considers task scheduling,energy optimization and time-varying electricity price jointly.In addition,the dynamic frequency adjustment of the processer is lack of consideration.In order to reduce the electricity cost in a green way,this paper simultaneously takes task flow modeling,energy optimization,task scheduling,and electricity cost minimization into account.The specific research contents include the following parts:First,for the data center system with delay-tolerant task flow,an electricity cost minimization problem of data center powered by smart grid is studied,and an energy-saving task scheduling strategy is then proposed.In this system,tasks submitted can be intelligently distributed to different computing nodes for processing.The data center can be powered by both conventional energy and renewable energy where the solar energy is considered in this paper.In order to cope with the uncertainty of renewable energy generation,we propose a flexible uncertainty model to describe it.After deriving the reference distribution of renewable energy generation based on prediction and historical data,an uncertainty set is defined to allow the real distribution of renewable energy generation to fluctuate within a certain range of the reference distribution.Then,a set of chance constraint approximation and robust optimization method is proposed to convert the uncertain renewable energy generation into a certain variable and solve the electricity cost minimization problem.Finally,two different task scheduling strategies are evaluated based on real world data.The effectiveness of the energy-saving task scheduling strategy proposed in this paper is verified,and the influence of various parameters on electricity cost and energy consumption is analyzed.Second,based on the above-mentioned uncertainty model on the renewable energy generation,the task scheduling and electricity cost minimization of the data center system with delay-constrained task flow are further studied.In this system,tasks can be intelligently distributed to different servers in different data centers where the delay constraints of tasks are described by the M/G/1 queuing model,and each server is equipped with a dynamic voltage and frequency scaling chip.Different data centers have different electricity prices and renewable energy generations.In order to solve the mixed integer nonlinear electricity cost minimization problem,we firstly reduce its complexity according to Jason's inequality theorem.Then a two-stage optimal server activation configuration strategy and task scheduling algorithm based on bisection method are proposed to obtain the global optimal solution.Finally,two different server configuration strategies are evaluated based on real data simulation experiments.In addition,the effects of different parameters on electricity cost and server activation configuration are analyzed,where the effectiveness of dynamic voltage and frequency scaling technology on server frequency adjustment is verified as well.Lastly,we summarize the energy optimization and task scheduling strategies of the above two data center systems,and discuss the future research directions.