Research On Energy Efficient Replication Management And Task Scheduling Under Cloud Environment

Energy efficient data management has already been paid much attention to byacademe and industry, since the energy consumption has become an important factor inthe further development of IT technology. In the life of a data center, the energyconsumption has exceeded the cost of hardware, and become the most important factor,except the cost of human resources. The burst of massive data promotes the developmentof cloud computing, which can be seen as a centralized management and distributedcomputing system with capability of massive data storage and computation. Challenge ofenergy comsuption, raised by the development of cloud computing and the increasingscale of clusters, has been paid more and more attentions by the researchers.This paper focus on energy efficient data replication management and taskscheduling under cloud environment. Based on marking computing node, anenergy-efficient data storage method and a task scheduling trategy are proposed.Togetherwith another low-power data storage method and a task scheduling trategy based on corearea model. Finally, evaluate the methods using Hadoop, an open-source cloudcomputing prototype proposed by Yahoo! The major work and contributions are asfollows:(1) Analysis the existing technologies of energy efficient data management, thedevelopment of cloud computing. A marking method of nodes in cloud is designed, basedon which propose an energy efficient data replication management method SNF-D(smallest node first-data replication) and an energy efficient task scheduling methodSNF-T (smallest node first-task scheduling). According to the routing environment, eachcomputing node is assigned a global unique code, which will replace node’s IP address toparticipate system task allocation. It can be proved that the mapping between nodes No.meets data integrity. The DVS based task scheduling method reduces energyconsumption of idle nodes, so as to achieve energy efficiency of overall system.(2) The core area model based energy efficient data storage method under cloudenvironment does not change the default data distribution, according to the data blockand the data node mapping relationship, selects " core area key node set" which meets thedata completeness. Base on this model realizes an energy efficient copy managementmethod and task scheduling strategy. The “core area key node set” contains all the data ofat least one backup, while the closure of not core nodes approach can effectively reducethe energy consumption of the whole system, to achieve the ultimate global energyefficient.（3）Set up the experimental system platform, by using Yahoo open-source Hadoopplatform.Under this test environment, achieved nodes marking based energy efficientcopy management and task scheduling method and core area model based energyefficient copy management and task scheduling method respectively. According toexperiment result, compared the methods proposed with the default processing method,and the results indicate that the proposed method gives energy efficient performance. Overall the methods proposed in this paper both achieve global energy efficient.