Research Of Grid Resource Scheduling Algorithm Based On Energy Optimization

Grid computing is a kind of Internet computing which has been emerging in recent years.Grid computing research aims to constructing a dynamic virtual organization in a network resources environment of distribution,heterogeneousness and autonomous dynamic,as well as achieving cross-organizational resource sharing and interwork,thus meeting the requirement of the comprehensive Internet-oriented applications for extensive computing and massive data processing.Traditional resource-scheduling algorithm can achieve these objectives effectively,while with the Ad Hoc networks and wireless sensor network(WSN) developing rapidly and being used widely in grid system,the problem of energy consumption is taking a more and more important place in grid resource scheduling.For example,on the Ad Hoc network,with the reserving energy for each resource node limited,the energy constraint has become a factor that can not be ignored in the resource scheduling of Ad Hoc grid.In addition,data-intensive grid applications has been increasing in recent years,such as high-performance research in physics,astrophysics research, weather forecasting and so on,which are based on the high-performance data grid with the characteristics of large energy consumption and long computing duration. Therefore,it has become an urgent problem in the field of grid resources scheduling to seek a new scheduling algorithm based on energy constraint.This thesis mainly studies two QoS constraints:energy constraint and deadline, and proposes a grid resource scheduling model based on energy optimization,on the base of which an improved algorithm is proposed.The main research and innovation of this thesis include the following areas:Firstly,the energy resources are added to the resource scheduling model because of the diversity of resources in the heterogeneous environment.The resources of traditional resource scheduling algorithm include computing resources,storage resources,network resources,hardware resources,etc,among which energy resources is rarely referred to,while it has become more and more popular in the grid. Secondly,a grid resource scheduling model is proposed from the point of energy optimization by considering energy consumption in the resource scheduling.The energy consumption is considered as a kind of QoS constraint and divided into communication consumption and computation consumption in this model.Energy consumption optimization is the most important target to be studied in this thesis.Thirdly,the constraint of Makespan is also considered as well as energy constraint and a model based on Makespan optimization is proposed.The Makespan optimization can solve the problem of load unbalancing caused by energy optimization and prevent any heuristic from mapping subtasks only to slow resources which consume less energy to execute a subtask.Finally,a resource scheduling algorithm(E_sufferage) is proposed based on the Makespan optimization and energy optimization.The goal of energy consumption optimization in the heterogeneous gird environment is achieved in the algorithm by considering the cost of time and energy.Gridsim toolkit is modified according to the proposed algorithms.Then the E_sufferage algorithm is simulated in the modified GridSim,which confirms the correctness and feasibility of the algorithms in practice.Overall,the main research and innovation of this thesis can solve the problem of the energy consumption in the resource scheduling.E_sufferage algorithm has realized energy consumption optimization,and better resources utilization compared with traditional heuristic scheduling algorithm.This thesis is supported by National Natural Science Foundation(60402028, 60773211),HuBei Natural Science Foundation(2008CDB335) and Supported by Program for New Century Excellent Talents in University(NCET-08-0806).