Research On Service Computing Offloading Technology In Fog Radio Access Network

The number and distribution of intelligent terminals show a rapid growth trend,and the user's service needs are more emphasis on experience.However,the current radio access networks have not yet made the ideal condition of typical services with intensive computing and ultra-low latency characteristics.The fog computing based radio access network(F-RAN)is proposed over a range of radio access networks,which further integrate the concept of fog into radio access network.First,starting with the development and characteristics of radio access networks,this thesis analyzes three network architecture modes:centralized MCC(Mobile Cloud Computing,MCC)mode,MEC-based(Mobile Edge Computing,MEC)architecture mode and based on F-RAN architecture model.Furthermore,by abstracting the system model of F-RAN for computing offloading of typical services,the thesis proposes the network model,communication model and computing model of a typical service under the F-RAN network.Moreover,the thesis analyzes the performance balance factor of computing offloading of a typical service under F-RAN in three aspects:computing cost,communication cost and application performance.Then,considering the difference in the performance level between the fog nodes,this thesis proposes two computing offloading models in F-RAN scenarios,which is used to systematically solve the problem of how to finish the typical service request of user equipment effciently.For the scenario where the fog nodes with sufficient resources can do at least one unit of typical service independently,the thesis proposes a dynamic programming algorithm based on surplus value to do computing offloading.The solution to three questions is given:the minimum service time cost,the computing task size that each fog node can allocate,the minimum number of nodes involved in the decision.It also verifies that the fog node can collaborate to complete the computation service with high computing requirements when the remaining resources are sufficient.Then,aiming at the situation that the fog node cannot complete a typical service under the limited resources,a greedy graph-coloring algorithm is proposed.The research gets the service-performance optimization model that multiple fog nodes collaborate to finish a typical service of a unit and give the solution of the minimum number of fog nodes involved in the computing offloading and the computational tasks that each fog node can be allocated.It is also verified that the fog node can also finish the typical service requirements with higher computational requirements through the cooperation of the fog nodes when the remaining resources are not sufficient.In the above two application models,the research puts forward and applies the concept of "surplus value",which is the performance of the workload of a fog node can offer.Finally,the research gets a better performance of the computation-offloading model by building a simulation platform.Here goes two experiments.For each experiment,there are three scenes for performance comparison:a single fog node,the average distribution and optimization model.