Research And Implementation Of Cloud Computing Resource Management Subsystem

Once cloud computing concept was proposed, because of the outstanding advantages of low cost and high efficiency of resource use it is welcomed by the major IT companies. The high reliability of cloud computing needs the mechanisms of security management and resource monitoring. The high scalability of cloud computing also needs the resource management system to support varieties of heterogeneous resources. The low cost of cloud computing services needs plenty of cheap PCs to be organized and needs efficient resource scheduling policy. So the cloud computing resource management largely determines the quality of service provided by the cloud computing platform and it is one of the key issues of the efficient and stable operation of the cloud computing system.In this paper, we analyze the main function of a cloud computing resource management system based on the cloud computing technology architecture and analyze the Map/Reduce resource scheduling model. In order to solve the local issues and the QoS constraints of this model, we propose three QoS-based resource scheduling algorithms. The algorithms support five-dimensions QoS constraints. Based on the analysis of the main algorithm design ideas used in the batch mode and online mode resource scheduling, we propose the load balancing resource scheduling algorithm under the guidance of the task QoS. The algorithm reduces the dependence of the integrated load-degree assessment on priori knowledge and reduces the sensitivity to the experience weight.In this paper, we analyze the research status of the cloud computing platform and resource management system. We design and implement the cloud computing resource management subsystem with reference to opensource cloud computing solutions Hadoop, in the context of the VCCP development. The subsystem is independent and implements the plug-in of resource scheduling strategy which is convenient for the updating and extension of resource scheduling algorithm. The subsystem also implements the resource registration cancellation which is convenient for the extension of heterogeneous resources. The tests show that the subsystem can meet the requirements of the virtual cloud computing platform resource management. In order to solve the subsystem performance bottleneck on the resource monitoring and the management tasks migration after the failure of the management node, we propose the distributed resource management system structure. The tests show that the distributed structure solve the lack of single-node management structure properties as well as the problems of management tasks recovery and migration in the central node downtime, to some extent.