Research On Dynamic Offloading And Management Mechanism Of Mobile Cloud Computing

With the rapid development of mobile applications and the increasing cloud computing service for mobile users,mobile cloud computing(MCC)was introduced as an integration of cloud computing and mobile environments.MCC offloads mobile terminals' computational and energy intensive applications to the cloud computing center through the mobile Internet,effectively reducing terminals'energy consumption,enhancing terminals' data storage capabilities,and extending terminals' battery lives.As one of the key candidate technologies of the 5th generation mobile networks(5G),MCC has been widely studied.However,the existing results are still not enough,and the perfonnance in terms of bandwidth,quality of service(QoS),mobility management,resource management,delay,and security needs to be further improved.Many key issues such as improving the quality of cloud services from the perspective of user's subjective experience,efficient bandwidth allocation,using cloudlet to reduce latency,improving service quality,providing seamless mobile cloud services,and developing security mechanisms by analyzing mobile user behaviors need to be addressed.Solving these problems has very important significance for the future implementation plan and technical support of 5G,and is also the basis for the selection of topics in this dissertation.This dissertation use mobile cloud computing as an entry point,focusing on multi-level resource management and finer granularity resource management operations under the constraints of dynamic network environment and terminal performance,bandwidth,and application features.The main contents of this dissertation are described in detail as follows:First,for a two-tier(cloud server-mobile terminal)mobile cloud computing architecture,a two-stage Stackelberg algorithm that considers spectrum efficiency,user QoE and cloud service provider's price to improve system performance is proposed.This algorithm can maximize network revenue under the guaranteeing system performance and user QoE.This algorithm also balances the performance of the terminal and the cloud,finds out the resource allocation strategy when maximizing terminal revenue and cloud revenue.In this dissertation,the impact of the user QoE and the prices provided by multiple cloud service providers on the system performance is analyzed,and the maximum value and minimum value of price are further calculated,and correctness of the algorithm is verified by simulation.Second,for the multi-tier mobile cloud computing architecture,the resource allocation problem of the mobile terminal,cloudlet and remote cloud service providers is modeled as a three-stage Stackelberg game.Consider the mobile terminal's bandwidth and power allocation strategy at the terminal side,the computing resources at the cloudlet side,and the pricing strategies at the cloud service provider side,the best response functions of bandwidth and power allocation,computing resource and price are designed.The existence of Nash equilibrium and the maximum and minimum price of the three-stage Stackelberg game are studied.The effectiveness of the three-stage Stackelberg game algorithm is verified by simulation.Third,under the joint consideration of mobile cloud computing and ultra-dense networks,the issue of how ultra-dense network users dynamically select cloudlet based on their own performance and costs is modeled as an evolutionary game.Cloudlet finds the best price and bandwidth allocation strategy based on the users' requirements.The problem of cloudlets' resource allocation is modeled as a non-cooperative game.In order to better solve the cloudlet selection problem and the resource allocation problem,these two problems are modeled as heterogeneous evolutionary game.In the evolutionary game model,the stability and uniqueness are proved.Moreover,an iterative algorithm based on population evolutionary game and Q-learning iterative algorithm based on reinforcement learning are proposed to obtain the equilibrium point.Among them,the algorithm based on Q-learning can enable users to obtain the optimal Nash equilibrium according to historical decision information,and solve the problem that it is difficult to establish a population evolutionary game management center in real life.Fourth,for the security issues of MCC network,it is very important to study user behavior because they are the basis of mobile cloud computing's safe and reliable operation.User behavior can be divided into altruism,selfish,and malicious.By analyzing user behavior,cloud service providers evaluate user behavior,correctly guide behavior,and formulate resource management mechanisms to make full use of the limited resources of mobile cloud computing.In this dissertation,resource management based on user behavior analysis is modeled as dynamic Bayesian game.Specifically,the cloudlet maximizes its own utility by setting a reasonable resource price according to the user's behavioral probability information.Malicious users can dynamically adjust their strategies based on their own performance and costs.We study the existence of Bayesian Nash equilibrium.The best response resource allocation functions for selfish and malicious users are obtained respectively.