Research On Profit Optimization-and High-Reliability-Configuration Of Multiserver System For Cloud Services

Cloud computing is an impressive business paradigm that leverages Internet to provide users with diversified or customized cloud services(e.g.,scientific computing,ecommerce,and graphic image processing).Cloud computing virtualizes infrastructure resources into common commodities for users to purchase on a pay-as-you-go basis,and users do not need to care about hardware-related upgrades and maintenance.In addition,with the rapid development of virtualization technology,and the increasing scale of cloud infrastructure deployments,users are easier to access and enjoy cloud services.As a result,cloud service providers such as Amazon,Google,and Microsoft Azure are becoming increasingly popular.In a cloud business market,service providers have a natural goal of maximizing profits while providing cloud services to their customers.At the same time,users are more concerned about service experience when accepting cloud services,that is,obtaining highly reliable cloud services in terms of timeliness and correctness,and then they will be more willing to trust and purchase cloud services.If cloud service providers want to maximize the service profit,they must consider the corresponding needs of users.Therefore,how configuring a high-profit and highly reliable multiserver system according to the characteristics of cloud service requests is particularly important for cloud service providers.In the cloud business paradigm,maximizing the service profit is both the goal for cloud service providers to configure multiserver systems for users and also the potential driver for the continued development of cloud computing technologies.The revenue from cloud services and the investment in infrastructure resources are the two main factors that affect the profit of cloud service providers,and both of these factors are highly dependent on the configuration of multiserver systems.This is because multiserver system configuration determines whether subscribers receive reliable and high-quality cloud services,which in turn is necessary for cloud service providers to increase revenue.At the same time,the multiserver system configuration is also closely related to the cost of hardware resources,which are used by the cloud service provider.Most of the existing research on the configuration of multiserver systems is to optimize the configuration from the perspective of energy saving,and due to the limitations,these works cannot be applied to the cloud computing business paradigm where cloud subscribers chase high-quality cloud services and cloud service providers chase profit maximization of cloud services.In addition,when targeting multiple cloud service application domains,i.e.,for different cloud service application domains,there is a need to configure differentiated multiserver systems since a multiserver system is devoted to serving one type of application domain.The budget of the cloud service provider is generally limited,and it cannot afford to invest in all application domains.Hence,the budget constraint is also a problem that cloud service providers need to concern.To address the above issues,this paper develops models that can capture the dynamic demand of the cloud service market based on user satisfaction and user-perceived value,and fully considers the impact of the heterogeneous characteristics of cloud service requests on timeline requirements,deadline miss rate,and reliability on multiserver configurations,and designs multiserver configuration schemes including customer perceived valued-and risk-aware multiserver configuration for profit maximization,deadline and reliability aware multiserver configuration optimization for maximizing profit,and the budget constraint and reliability enhancement multiple multiserver systems for maximum profit,respectively.The specific contributions of this dissertation are as follows,1.This dissertation addresses the problem of configuring multiserver systems from the perspective of optimizing cloud service provider profits and proposes a userperceived value and risk-perceived multiserver system configuration scheme based on the dynamic pricing of cloud services.The scheme develops a user-perceived value model for multiserver system configuration for cloud services based on the requirements of the average scheduling policy using a dynamic pricing strategy based on user satisfaction,using which the dynamic demand of the cloud service market can be captured.In addition,in this scheme,this dissertation proposes a feedback control mechanism to suppress the possible profit-risk in dynamic pricing.In this study,this dissertation constructs a cloud service model and a perceived value model based on user satisfaction and uses the developed perceived value model to further build a dynamic demand model for the cloud service market,and finally defines a multiserver system configuration problem for profit optimization of cloud services based on the constructed model.To solve the problem,this dissertation not only presents an analytical method to solve the optimal multiserver system configuration under the problem definition but also develops a heuristic method based on simulated annealing to quickly find the suboptimal but effective multiserver system configuration.2.This dissertation addresses the problem of configuring multiserver systems from the perspectives of optimizing cloud service provider profits,cloud service deadline miss rate,and multiserver system reliability,and proposes a deadline and reliabilityaware multiserver system configuration scheme based on the heterogeneity of cloud service request timeframe requirements.The scheme develops a soft-error reliability calculation model for multiserver systems oriented to average scheduling according to the requirements of the average scheduling policy and derives a deadline miss rate calculation model for cloud service requests with soft real-time requirements in conjunction with fault tolerance techniques for re-execution.In this study,this dissertation constructs a cloud service model,a multiserver system reliability model based on fault-tolerance techniques of re-execution and a cloud service deadline miss rate model,and finally defines a multiserver system configuration problem that optimizes the cloud service profit under the system service rate,multiserver system size and server rate constraints on the basis of the constructed model.To solve this problem,an iterative method based on the Lagrange multiplier method is designed in this dissertation,which can effectively find the optimal multiserver system configuration and the corresponding maximum cloud service profit under the definition of this problem.3.This dissertation addresses the problem of configuring multiserver system configurations for different cloud service application domains and budgets from the perspectives of cloud service application domain selection,budget allocation,and multiserver system reliability,and proposes a multiserver system configuration scheme for reliability enhancement in multi-application domains under cost constraints based on the heterogeneity of cloud service request timelines requirements.The scheme proposes a computation model for the expected revenue of cloud service requests with different slack based on the requirements of the average scheduling policy,redesigns the multi-server system reliability computation model,and considers the impact of multiserver system reliability on the system service rate.In this study,this dissertation constructs a cloud service model for multiple application domains,an expected revenue calculation model for cloud services with different slack,a multiserver system reliability calculation model,a system service rate model,and a cloud service application domain selection model,and finally defines a profitoptimized multiserver system configuration problem in multiple cloud service application domains on the basis of the constructed model.To solve the problem,this dissertation proposes a single application domain optimal configuration scheme based on the penalty function method and cleverly designs an application domain selection strategy based on the backpack problem using the single application domain optimal configuration scheme to maximize the cloud service provider’s cloud service profits.