Research On Dynamic Service Dependability Assurance Models And Methods In Cloud Computing Environments

The advent of cloud computing has profoundly changed computing modes in various fields such as military,transportation,industrial manufacturing,and social services,allowing people to gain on-demand computing power without buying large quantities of expensive specialized equipment.With the widespread adoption of cloud computing infrastructure and next-generation communication technologies,the concept of "anything as a service" is becoming a reality,where people can get services from clouds anytime and anywhere.However,such technical developments may also lead to various attack methods.Due to unreliable links,malicious nodes,and vulnerable shared cloud services,it is challenging for users to obtain reliable cloud services.Hence,it is urgent to study service assurance in cloud computing environments.Since resource provisions,service selections,service combinations can be from various remote cloud computing environments,virtualization mapping methods make these links highly dynamic that it is difficult to use static assurance methods to obtain ideal results.Therefore,this dissertation proposes a dynamic service assurance method in cloud computing environments under untrustworthy network conditions.The method makes comprehensive use of various means to form a feedback process,including screening of cloud computing resources,selecting service attributes,performing service combination and computing power re-optimization,which can dynamically guarantee that users obtain trusted on-demand services.Firstly,to address the need for dynamic assurance of trusted services in cloud computing environments,a dynamic assurance model of trusted services for cloud computing environment is proposed.The model integrates a multi-layer feedback control mechanism for cloud computing resource filtering,service attribute selection,service combination execution and computing power re-optimization,which can guarantee service dependability dynamically at multiple levels.Due to the fundamental role of resource virtualization in cloud computing systems,subsequent assurance methods will be impossible to achieve if computing resources are unreliable.To address the limitations of unreliable virtual resource mapping,a mapping algorithm SAPSP-VNE is proposed based on improving the simulation of annealing particle groups.With a goal of minimizing the underlying network resource overhead of virtual network mapping,the problem of virtual network resource allocation of virtual network dependability is converted as an integer linear planning model.This model can ensure that the rest of the virtual network can still have connectivity and maximize the dependability of the virtual network when underlying physical networks fail.Experimental results show that SAPSP-VNE algorithm can effectively reduce the average utilization rate of resources,and improve the success rate of mapping and the recovery success rate of virtual networks.Secondly,to address the complex problem of service feature selection caused by the high dimension of web service features in cloud environments,a service feature assessment and subset selection method ISVM-FSM is proposed,which is based on entropy and SVM model.The algorithm uses fuzzy centralized information entropy to calculate the distances between the vectors in each category using the scale factor of the feature and the entropy value between specific categories,which can solve the over-complex issue caused by large feature spaces.When solving the subset of service features,the method leverages the idea of forward distribution algorithm,which can obtain a higher accuracy than the traditional SVM model by constantly fitting the residuals of training data in previous model.Experimental results show that this method can effectively select important features from service data sets with redundant features,and the scale of subset of features is significantly compressed and the accuracy is improved compared with traditional algorithms.Thirdly,to address the limitations from third-party platforms on traditional service combination methods and the requirements of trusted web service diversity are difficult to meet,a smart contract service combination method based on blockchain is proposed to build a secure and trusted cloud service trading environment.Based on blockchain-based service smart contracts,a hybrid grey wolf optimization algorithm HGWO is proposed as the core algorithm of service combination for dynamic service scheduling optimization.The proposed algorithm overcomes the shortcomings in traditional meta-heuristic methods,such as low solution accuracy,slow convergence in later stages,and the tendency of falling into a local optimal problem.Experimental results show that HGWO algorithm has better convergence,which can improve the accuracy and diversity of solutions,providing effective solutions for service combination in cloud environments.Finally,to optimize computing power while providing trusted services,a cloud computing service optimization scheme based on service migration is proposed.In this method,the trusted service optimization problem is converted as a problem of minimizing system overhead,where the resource states of cloud nodes and central cloud are modeled.On this basis,a resource optimization strategy is obtained by improving the genetic algorithm.Experimental results show that the proposed scheme can reduce the overhead and effectively improve the performance of the system.