Resource Allocation Approach With High Energy Efficiency In Cloud Computing Environment

Cloud computing technology already has a really high maturity at present. Along with the growth of cloud infrastructure services and the unceasingly expansion of cloud computing scale, the energy consumption of IT resources surge, which has greatly hindered the development of IT industries. Energy consumption has become a new reason that restricts the development of cloud computing. In this context, how to reduce the energy consumption of cloud computing has become a new problem.Energy efficiency is the effectively task load with per energy conservation. For a task, the less of energy consumption, the higher energy efficiency. The resource allocation problem is the key technology in cloud computing. Most job schedulers focus on the job scheduling, not considering the allocation of resources and energy efficiency. So, a resource model is needed to forecast the resources needed by tasks, raising the resources utilization and energy efficiency of the cluster.The concept of awaiting energy consumption of resource and resource ratio is proposed and proved in this thesis. This thesis argues that allocating resources based on resource ratio is the effective means to improve energy efficiency on software level. This method is innovative. This thesis defines the energy consumption for resources waiting as Awaiting Energy Consumption of Resource, defines the ratio of resource quantity that makes there is no awaiting energy consumption as Best Resource Ratio of Task, defines the process of allocating computing resources and storage resources for the Map/Reduce tasks as Resource Allocation. First of all, the general resource ratio model is put forward in this thesis after studying the characteristics of different resources possession. Then, based on the MapReduce programming model, Map/Reduce task specific resource ratio model is given by analyzing the resources possession features of Map/Reduce tasks. Based on the model, a high energy efficiency resource allocation approach R2 is proposed at the last, which include R2 scheduler and R2 resource allocation.Finally, some experiments are done based on the Hadoop environment. First, the resource ratio and best resource ratio is proved to exist. Then the Map/Reduce task division given in this thesis is proved to be reasonable. Thirdly, by comparing with the experimental data, showing that R2 approach can significantly reduce the awaiting energy consumption of Map/Reduce task, and improve resource utilization of each node at the same time.The resource allocation approach proposed can be applied to current MapReduce systems, And it has certain theoretical significance and practical application value for energy efficient cloud computing model and the optimization of energy efficiency.