| Cloud computing is a mature and developed model that can provide resources and services to many customers at the same time.These resources are provided in the form of virtual machines that are created for the user and allocated on the physical server.These services are provided when the customer asks for,and are performed by the service software components that reside on the cloud server.In order to improve the viability and fault tolerance of the most critical service components,the service provider uses redundant service components to compute the same task.N-Version Programming(NVP)based on voting is one of the popular redundancy techniques.Using voting to obtain highly reliable data from several different versions of service components can improve the reliability of the cloud network.But at the same time,in a cloud computing environment,the virtual machines of different users usually run on the same physical server and are logically isolated from each other.However,malicious users can bypass logical isolation and obtain sensitive information from virtual machines co-residing on the same server,thus forming coresident attacks and affecting the security of cloud networks.The research work of this thesis is as follows:(1)Under the premise of co-resident attacks in the cloud environment,the theoretical basis for cloud network to use NVP technology as a means to improve fault tolerance ability is explained.On the basis of this theory,the security and reliability of the small cloud system is analyzed,and guided by minimizing the loss of cloud service provider,the number optimization model of service components is established with resource constraints.(2)Based on the hardware fault data of the cloud physical server and the real-time data under stable pressure,the Agglomerative Hierarchical Clusering algorithm was improved to a clustering algorithm integrated with three fault domains,and the fault prediction model of the cloud physical server was trained.The accuracy of the model is illustrated by the test data.(3)In the cloud network with high reliability requirements,this thesis proposes to use disguise components under resource constraints to improve system security when facing co-resident attacks.Furthermore,guided by maximizing the probability of service success,the fault prediction of cloud physical server is taken as the basis to decide the optimal deployment strategy of NVP service based on FFP voting under the resource and time constraints.(4)Given the example and sensitivity analysis of parameters it is shown that these strategies can effectively help NVP service components to resist the co-resident attack in small-scale cloud network system.By doing fault prediction and modeling the NVP service under the threat of coresident attacks,the solution of optimized NVP service presented in this thesis has great significance for the intelligence of cloud computing. |
