Energy Consumption Optimization Oriented Data Center Resource Dynamic Scheduling Model And Solution

With the development and progress of social economic and technology, therelationship between IT (Information Technology) industry and human being work andlife becomes closer. The fast growth of IT industry enhances more and more data centers,but the energy consumption also turns to a more serious topic with the service by thesenumerous data centers. Energy conservation is one significant issue nowadays, and it isone of the long-term strategic tasks in our country. The great increase of energyconsumption in data center already becomes the most serious threaten for power supplycapacity. Hence, the energy conservation of data center turns to the important problem tobe solved at present. In this dissertation, we focus on the power consumption resourcedevice of data center. For the various types and huge scale of resources in data center,and the requirement to satisfying the uncertainty of customers, the failure of control orscheduling for resources would waste large amount of power. Therefore, this work focuson the energy consumption optimization oriented data center resource dynamicscheduling problem in this dissertation.Firstly, we review and analyze the recent research works about the data centerresource scheduling problem, propose the future research direction, and introduce in briefthe simulation software for data center resource energy consumption.Then, we present the data center resource models, and discriminate the cyberresource and physical resource according to the resource use. Based on the analysis foroperational principle of main energy consumption device, we present several basicmodels for multiple types of resources, and describe the services provided by the datacenter resources and the relationships among these resources. According to thecharacteristics and applications of resources, we present the energy consumption modelfor the data center, and propose definition of the data center resource dynamic schedulingproblem. Focused on the characteristic of data center resource scheduling, we present theCyber-Physical System based two layers scheduling model.Due to the characteristics of multi-type and multi-attribute or data center resources, in this dissertation we propose dynamic programming based scheduling solution. Wedescribe the basic concepts, main idea and solving process in detail firstly; then, presentthe data center resource dynamic scheduling model based on the idea of dynamicprogramming; next, we propose the heuristic rule and the dynamic programming basedsolution for the problem. Finally, we take experiments for the training stage for thealgorithm.The work in this dissertation integrates the simulation software of EnergyPlus andGreen Cloud Simulator for data center resource energy consumption. We simulate thetraining process on this integrated platform to validate the solution, and test theadvantage of the dynamic programming based scheduling algorithm through thecomparison with other scheduling solutions.