Key Technology Of Service Dependability Assurance In Distributed Virtualized Environment

With the rapid development of network communication technology and software development mode,new forms of large-scale complex computing continue to emerge in large numbers,and XaaS has become the ultimate goal of the future development of general-purpose computing environment.Based on distributed virtualization technology,distributed virtualized environment combines the abilities of resource integration and cross platform,achieving the efficient management and on-demand allocation of highly heterogeneous and autonomic computing resources.DVE can guarantee to meet the users' diverse and on-demand requirements,which is driving and supporting the rapid change of computing form towards the service-center system,so that XaaS gradually become a reality.Because the distributed virtualized environment is much more open,sharing,dynamic and transparent than the traditional one,new challenges have been brought into each stage of the construction process of service compositions.A series of dependability assurance problems,such as finding a service component to obtain high dependability assurance in dynamic transparent environment,improving the dependability assurance during the dynamic process for constructing service compositions and continuous evolution of dependability enhancement in the running process of service compositions are arise with the development of distributed virtualized environment.Accordingly in this paper,service dependability assurance in distributed virtual environment is studied.Based on the theory of evolutionary game,semantic web,machine learning,intelligent optimization and other related theories,a dependability assurance model of service behavior is established,and full period dynamic guarantee for service dependability is achieved by studying the key technologies of the service dependability assurance in different stages of the service composition process.The main contributes of this paper are as follows:Firstly,based on the dynamic evolutionary game theory,a multi-strategies model of service behavior dependability assurance is proposed proceeding from two levels of service behavior definition which is service behavior compatibility and strategic evolution.We analyze the formal description of service interaction by Petri Net,deploying thecorresponding strategies to deal with different types of service interface incompatible in order to solve the problem of service behavior compatibility,which causes the insufficient of dependability assurance.On the other hand,according to the different trust level of service entity behavior,service behavior strategies are graded.Multi-strategies trust evolution based on Moran process is used to optimize the service strategy and enhance the dependability of the service system.Secondly,in order to cope with the phenomenon of False Positive and False Negative in the discovery and matching of existing semantic services,an adaptive hybrid semantic service matching method based on random forest is proposed.By introducing approximate logic matching,precondition effect matching and other methods,the failure of the match caused by too strict logical matching is avoided with the help of logic concept contraction and abductive.Furthermore,different matching results are obtained as eigenvalues,the classifier machine learning is adopted by feature matching results using random forest method.The adaptive hybrid semantic service matching avoids the problems and shortcomings of traditional threshold dividing method.Thirdly,in order to realize the construction of high dependable service composition,a method for selecting trusted service components in distributed virtualization environment is proposed.According to the social nature of the communication process of service components,the concept of Quality of Dependability(QoD)is established to describe the dependability parameter.Based on the subjective evaluation of the trust of the service components and the objective attribute parameters such as the component intimacy in the distributed virtualized environment,a dependable utility function based on QoD is established.Using the intelligent optimization search algorithm,the optimal service path under QoD-constrained conditions is obtained.Then the service components on the optimal service path are selected to realize the dynamic composition of services,improve the assurance of dependable construction of service composition.Finally,in order to solve the problem of consistency guarantee in the process of dynamic evolution of service composition,a service substitution method based on transaction compensation is proposed.The transaction granularity and transaction scope are analyzed from the view of data flow between service components by taking full account of the transaction attributes of services.The compensation mechanism supports transaction levelservice substitution,maintaining the consistency of service composition in the substitution process.It avoids the reduction of dependability after failure-handling,and provides supports to achieve service composition dynamic enhanced evolution.