Concerning the complex structure and huge amount of data of resource management in cloud computing, a package-cluster mapping based resource management framework was proposed. In this framework, resources are allowed to be shared in a package among virtual machines, and the resource scheduling becomes more flexibly by using a specific resource sharing model. Moreover, an improved package-based Genetic Algorithm (GA) was used in this framework, which encoded package groups with chromosome group and resource pattern, and designed crossover operators and mutation operators according to the change of the chromosome length. The number of clusters and the resources of the packages were integrated, the scale of the problem was solved by using an abstract model. Experimental results showed that, compared with the traditional virtual machine centered framework based genetic algorithm and the package-cluster framework based adaptive algorithm, the CPU utilization of the proposed method was improved by 9% and 5% respectively, and the memory utilization was improved by 14% and 7%, respectively. It proves that the proposed package-cluster mapping framework based algorithm can effectively reduce the number of the used cluster nodes and increase the resource utilization rate.

In cloud resource sharing service model, in order to realize the multi-attribute range query of cloud resources, an improved E-SkipNet network was proposed. Firstly, based on the traditional Distributed Hash Table (DHT) network SkipNet, data attributes were added to the setting of NameID and physical nodes were added to single attribute domain to support multi-attribute range queries in E-SkipNet. Secondly, on the basis of the original E-SkipNet network, the physical nodes were simultaneously mapped into multiple logical nodes and added to multiple attribute domains, and the resources were released in accordance with different attributes to different logical nodes in the improved E-SkipNet. Finally, the resources were mapped to logical nodes utilizing uniform locality preserving hashing function, which was the key to support efficient range query. The simulation results show that the routing efficiency of improved E-SkipNet network was respectively increased by 18.09% and 20.47% compared with E-SkipNet and Multi-Attribute Addressable Network (MAAN). The results show that the improved E-SkipNet can support more efficient cloud resource multi-attribute range queries and achieve load balancing in heterogeneous environment.