Optimal Pricing And Capacity Planning Strategy For A Monopolistic Cloud Service Provider Considering SLA Constraint

Cloud computing is a revolutionary paradigm of the next generation IT industry,which evolve out of established technologies,such as cluster computing,grid computing and virtualization.Cloud services make IT resources available to customers in a pay-as-you-go manner.Furthermore,service-level agreements(SLAs)are negotiation between the providers and consumers,which provide guarantees for service capacity and execution time.Decision on price and capacity of cloud service is an essential issue in cloud service operation management.Extant pricing strategies always consider short-term gains from the perspective of providers,and customers just passively accept the price.In addition,cloud service providers do not take full account of SLA impact on service capability planning within a decision-making process.The optimal strategy on long-run dynamic pricing and capacity planning for cloud service under SLA constraints is studied in this thesis.Market-oriented pricing and resource management is necessary to regulate the supply and demand of cloud service to achieve market equilibrium.As a consequence,a bi-level programming model is built to represent a global optimization process of cloud service,with the monopoly provider as the leader and all customers as followers.We analyses cloud service from the perspective of provider.At the first level,the profit is optimized for the monopolist by deciding service level to be offered to all potential customers,with its corresponding price.We discuss how cost can be classification and its change trend when analyzing cloud service operation process.At the second level,all the potential customers in cloud computing market make self-selection decisions on buying or not the cloud service for gaining non-negative utility.Customer's willing to pay as a piecewise linear function for heterogeneous service demands.Because the model consists of a complicated profit function and a piecewise linear utility function,getting an analytical solution is impossible.Thus,we use an improved algorithm is proposed here based on the combination of niching steady-state genetic algorithm and hill climbing method to compute numerical solutions of this dynamic pricing problem.The thesis investigates supplier's decisions affected by resource utilization rate,default rate,equipment distribution ratio and customer tolerance coefficient with one and two-dimensional customers utilities.Customers' utilities measure need considering response time,which composed by service time and waiting time.The results provide insights into four areas: optimal profit,pricing,capacity planning strategy,and supply-demand relationship.According to the characteristics of three basic service layers(Iaa S,PaaS and SaaS),some novel managerial implications are obtained.We demonstrated that provider's decision-making needs to combine economy characteristics and a resource optimization mechanism.The proper policies have a significant impact on cloud service market in a long term,which also has a certain practicality in industry.