Research On Node Clustering And Resource Allocation Algorithm In Fog Computing Networks

With the development of the Internet of Things(Io T)technology,more and more mobile devices are connected to the Internet.However,mobile devices have a problem that their computing resources and battery capacity are limited,and they cannot process tasks that need large computing resources.As a further expansion of Mobile Edge Computing(MEC),Fog Computing Network(FCN)deploys fog computing nodes at the edge of the network to provide additional computing and storage capabilities.Therefore,the mobile terminals can choose to offload tasks to the nearby fog computing nodes to reduce time and energy consumption of processing task.Quality of experience(Qo E)of application services can be improved greatly.This thesis focuses on the research of resources allocation algorithms among fog nodes in fog computing network.Proposing balanced clustering and joint resources allocation algorithm(BCJRA),and multi-convex inequality-constrained alternating direction method of multipliers(MIADMM)based distributed resources allocation algorithm(MIADMM-DRAA).The main work of this thesis is shown as follows:(1)Balanced clustering and joint resources allocation algorithm of the fog computing networks is proposed.The algorithm mainly consists of balanced clustering module and resource allocation module.The balanced clustering module makes use of the local characteristics of edge behavior of the fog computing networks,comprehensively considers the computing resources,workload and geographic location of fog nodes,and defines the loss function.The clustering method based on K-means obtains clustering result that minimizes the loss function,so that fog nodes in fog computing networks will form several clusters with balanced computing resources,workload,and distance between the fog nodes.After obtaining the clustering result,resource allocation module optimizes the task offloading matrix,computing resource allocation matrix and communication resource allocation matrix of each cluster parallel to reduce maximum delay and maximum energy consumption and achieve the cooperation among adjacent fog nodes.The results of computer simulation show that the computing scale of the resource allocation problem in the fog computing networks and the pressure of data transmission in the network is reduced after balanced clustering.BCJRA can quickly converge and effectively reduce the maximum delay and maximum energy consumption of fog nodes in the fog computing networks.(2)MIADMM-based distributed resource allocation algorithm in the fog computing networks is proposed.This algorithm takes the advantage of distributed architecture of the fog computing networks and achieves allocation of computing resources through the cooperation of fog nodes in fog computing networks to minimize network delay and energy consumption.MIADMM-DRAA develops the model of task arrival rate in fog computing network based on the M/M/1 queue model in queuing theory,derives the delay and energy consumption of fog nodes,and establishes an optimization goal that minimizes the total system cost.According to the number of fog nodes in fog computing networks,the algorithm divides the resource allocation problem into multiple sub-problems.Each fog node only solves one sub-problem and the base station only plays the role of global data exchange and broadcasting.Therefore,the network transmission cost is reduced.Simulation and analysis results show that MIADMM-DRAA has fast convergence speed and the algorithm can obtain lower network cost compared with other existing algorithms.