Optimal Decisions In Several Stochastic Service Systems Based On Queueing Game

In stochastic service systems,arriving customers may experience system conges?tion and queuing phenomena due to the randomness of arrival process and service pro?cess.The customers and managers in the systems will decide what strategies to optimize their expected net benefits according to their own interests.Ignoring the game between customers and managers will lead to the wrong evaluation of system performance and then make wrong decisions.From the perspective of queueing game,this thesis studies the optimal decision-making problems of stochastic service systems under the back-ground of cognitive radio system and web service system with self-supporting server and cloud service.Firstly,this thesis considers a cognitive radio system with a single primary user in different information level cases.When analyzing the game between users,the game between primary users(PUs)and secondary users(SUs)is regarded as the Stackelberg game,and the internal game among PU's service requests(or SUs' service requests)is a non-cooperative game.By combining Stackelberg game and Nash equilibrium anal-ysis,the strategic behaviors of PU and SUs are obtained based on different information disclosure levels.When users adopt Nash equilibrium strategies,from the perspective of game theory the performance measures of the system are discussed,and the relevant mathematical formulas are obtained.Theoretical and numerical analysis shows an in-teresting counter-intuitive phenomenon,that is,the expected sojourn time of SUs is not monotonously decreasing in the PU's service rate.Specifically,it is non-decreasing and then non-increasing in the service rate in the case without queue length information,and shows more variability in partial and full queue length information cases.Therefore,the increase of PU's service rate does not necessarily improve the system performance.Since the system often needs to guarantee the quality of service(QoS)for PU,this the-sis investigates and finds optimal service rates of PU in different information levels to meet the PU's QoS requirement and simultaneously to maximize SU's throughput from the viewpoint of the service providers.Subsequently,comparing preemptive service mechanism with non-preemptive service mechanism,this thesis finds that from the per-spective of throughput maximization,there exists a critical value,and if the maximum sojourn time that PU can withstand is less than the value,the preemptive service mech-anism is better,but from the perspective of social welfare,the non-preemptive service mechanism is always more advantageous.Secondly,for a cognitive radio system with two primary users,this thesis models it as a two-server preemptive pr:iority queuing system,deduces the formulas of perfor-mance measures,and analyzes the optimal pricing strategy from the perspectives of service providers and social managers.Based on the stochastic waiting cost and the deterministic waiting cost,this thesis deduces the optimal pricing strategies for maxi-mizing the revenue of service providers,and finds that providing free service to all users is socially optimal.Finally,this thesis also considers a web service system with a self-supporting serv-er and a cloud server.In this system,customers send their service requests to a web ser-vice company.Since the cloud service may be subject to security risks,a risk-sensitive web service company needs to establish admission control so that the arrival rate of ser-vice requests entering the cloud server does not exceed a threshold.When an arriving service request can not directly access cloud service due to the admission control,he will decide whether to join the internal subsystem or enter the orbit space to retry the cloud service with a certain retrial rate based on the available information.This thesis obtains the conditional equilibrium probability of customers joining the internal subsys-tem,and discusses the optimal queue length limitation and the optimal retrial rate from the perspectives of managers and opportunists,respectively.However,managers and opportunists are different interest groups,and the opportunists may be cooperative or non-cooperative.Based on the Stackelberg game,the joint optimal value of the queue length limitation and the retrial rate is got.Numerical analysis shows that the oppor-tunists always choose the same optimal retrial rate whether they are cooperative or not.In addition,the existence of opportunists will seriously damage the social welfare.