QoS-aware Service Resource Scheduling And Optimization

Services computing is a burgeoning interdisciplinary computing paradigm.It effectively devises,operates,manages and optimizes various information services and business services,through the ingenious design and utilization of computing resources and information technologies,which has been widely practiced in diverse fields.With the continuous expansion and evolution of services computing system,the number of service providers and users increases rapidly.Accordingly,the services computing system is gradually with more complicated compositions,which presents an ecological trend.Multiple users/service providers are collectively involved into the operation of the services computing system,and form a competitive situation.For users,they need to encounter the competition from other users for service resources,whose purpose is to win the optimal target service/resource allocation scheme with the high Quality-of-Service(QoS)level obtained.For service providers,they need to take an appropriate incentive strategy to stimulate the maximum users who purchase service resources,in order to meet the QoS demand for the maximum users.It helps to consolidate the maximum market share within the competitive services marketplace.For the services computing system,it necessitates a fair competitive environment for services resources,where service quality and service fairness should be jointly addressed.In this way,a benign competition order is built up within the ecological services computing system,which contributes to the systematic sustainability.To address these issues,this dissertation primarily focuses on the management of competitive situation between multiple users and service providers,within the ecological services computing environment.From three aspects(i.e.,service selection,resource pricing and resource management),the QoS-aware service scheduling and optimization is extensively investigated,where the specific QoS optimization models and solutions are proposed.The dissertation is summarized as follows.1.Fairness-aware multi-user service selection method is studied.In details,a QoS-aware multi-user concurrent service selection model with constraints is firstly formulated.Then,by means of introducing the Max-Min Fairness(MMF),the lexicographical optimization problem which achieves the MMF is hereby given.According to the uniqueness of MMF optimization problem,a Fairness-aware Concurrent Service Selection algorithm named FASS is proposed,which is the iterative Linear Programming.The optimality of FASS algorithm is theoretically proved,while the effectiveness and efficiency of FASS algorithm are sufficiently verified and evaluated through simulating experiments.2.Market-oriented cloud resource pricing mechanism is studied.Firstly,the cloud resource auction marketplace is formulated.According to the QoS-aware user utility model,the cloud resource purchase strategy based on individual rationality is defined for users.Then,the Price-Incentive Resource Auction mechanism named PIRA is designed to maximize the user incentive.It targets to stimulate the maximum users who purchase cloud resources,on the premise of the minimum profit-rate requirement by the cloud service provider.In addition,the PIRA mechanism is theoretically proved to satisfy the budget feasibility,incentive compatibility,and envy-freeness.These essential properties can enhance the robustness of the cloud resource auction mechanism PIRA.The actual performance of the proposed PIRA mechanism is extensively verified through simulating experiments.3.Distributed Quality-of-Experience(QoE)-aware task scheduling and resource management method is studied under the edge computing paradigm.Firstly,the edge resource allocation model is demonstrated and formulated,and then the correlative between QoS and QoE metrics is analyzed quantitatively.The Edge Resource Allocation problem named ERA is given with the objective of maximizing the systematic QoE level across multiple IoT users.To attack the research challenges of solving out the ERA problem effectively,the potential game model is introduced to make the ERA problem optimized in a decentralized manner.On the basis of the preemption-based QoE improvement mechanism and cooperative message-passing mechanism,the QoE-Aware Decentralized Edge Resource Allocation Algorithm named QoE-DEER is designed to obtain the Nash equilibrium with the highest systematic QoE state.Finally,the performance and convergence of decentralized QoE-DEER algorithm are theoretically proved and experimentally verified.