Virtual Resource Management For Cloud Service Provisioning

Cloud computing is a newly-developing calculation mode based on network service and application, which provides users with the scalable computing resource in a way of virtualization through managing, scheduling and integrating all kinds of the distributed resources in the networks. In the cloud computing, the core issue that users concern is to obtain the cloud service. Because the user is far away from cloud resources, there is not enough information to identify the quality and reliability for the service in the face of numerous service selections and multi-tenant sharing application service. Whether the cloud service can meet demands that constantly changed still remains a question.Concentrating on the service issue of the multi-tenant application in cloud computing environment, the research about service resource virtualization management was conducted in this paper. On the basis of the existing related technologies about cloud storage, the highly availability was guaranteed by optimization of redundant disaster preparedness for the data stored in the cloud.Cloud service resource was correctly classified according to its resource properties in the paper, and then cloud service resources were managed uniformly in a manner of virtualization. The main research included the following aspects:1. The existing cloud storage technologies were analyzed, and the existing research progress was summarized systematically. Meanwhile the key cloud storage technologies were discussed and the problems existing in cloud storage technologies were expounded in the paper.2. This paper proposed a cloud storage data redundancy scheme, which optimized the storage data redundancy using redundancy technology of erasure code, ensuring the data reliability. Then the Chord ring was employed to improve the retrieval efficiency of the distributed storage data block.3. A cloud service resources classification strategy was introduced in this paper, the naive bayes algorithm was improved based on feature similarity. And then the improved classification algorithm was parallelized using MapReduce programming model in the Hadoop cloud platforms. This strategy not only ensured the accuracy of classification, but also improved the ability that system deal with mass of data.4. A model for cloud service resource virtualization management was designed in this paper. It used an open source cluster system named PBS(Portable Batch System), in order to provide users with a complete and unified application interface; The Xen virtualization technology was used to manage PBS in a way of virtualization. It enabled the system to response the quest of cloud resources data in a faster way, improved the system resource utilization and reduced the cost of the hardware and software configuration.