The Research On Task Offloading And Resource Adaptation Of Edge Computing Service In Smart Identifier Network

With the development of Internet technology,network services and network applications have affected various fields.With the rapid increase of network data flow and computing tasks,it is difficult for the traditional network system to meet the demand for efficient,reliable,massive and ubiquitous services.To solve the problem of triple binding of traditional Internet,the Smart Indentifier Network(SINET)proposed the architecture of "three layers and two domains",which provided a new idea for realizing the manageability,controllability,openness and flexibility of the network.Edge computing(EC)can provide better quality of service by using edge computing resources for computing services in Smart Indentifier Network.However,the service complexity and resource limitation of edge networks bring new challenges to the dynamic and flexible control of computing services in different scenarios.In this paper,according to different demand characteristics and optimization objectives,we study the task offloading and resource adaptation control optimization strategy based on Smart Indentifier Network for terminal complex computing service,terminal streaming computing service,multiterminal competitive computing service and edge convergence computing service in edge network.The main work and innovation of this paper are introduced as follows:(1)Aiming at the problem of task offloading and resource allocation of terminal complex computing services in edge network,a joint management and control optimization strategy of complex service segmentation and partial task offloading is proposed.In this paper,the optimization objective is to minimize the processing overhead of complex computing services.By comprehensively considering the characteristics of service model,task dependency,node device capability and wireless channel status,a terminal complex computing service system model and control mechanism based on SINET and EC are proposed.A joint optimization model of complex service segmentation and partial task offloading is constructed.Based on deep learning method,a service computing model for video stream body posture estimation is built.By analyzing the computational load and data flow of the constructed model,using neural layer grouping and pipeline processing,a Threshold Particle swarm based Collaborative Partition Offloading(TP-CPO)algorithm is designed.The performance is evaluated by simulation experiment.Simulation results show that the proposed strategy can effectively reduce the service response time and terminal energy consumption under different channel bandwidths and collaborative server loads.(2)Aiming at the problem of task offloading and resource allocation of terminal streaming computing service in edge network,a joint management and control optimization strategy of task offloading scheduling and terminal power control is proposed.This paper proposes a joint management and control strategy of task offload scheduling and terminal power control.This paper takes the task processing utility maximization of streaming computing service as the optimization goal.By considering the arrival task characteristics,terminal energy state,wireless channel status and server load,the model and control mechanism of terminal streaming computing service system based on SINET and EC are proposed.The joint optimization model of task offloading scheduling and terminal power control is constructed.According to the Markov property of the terminal waiting computation task,the constraints of offload scheduling decision and power control decision are analyzed.The state,the action,the reward,and punishment functions are designed.A Hierarchical Deep Reinforcement Learning based Adaptive Scheduling Control(HDRL-ASC)algorithm is implemented.The parameter tuning and performance evaluation of the algorithm are carried out through simulation experiments.The simulation results show that the proposed strategy can effectively improve the task processing efficiency and reduce the terminal power consumption under different task arrival rates,wireless channel status and server computing capacity.(3)Aiming at the problem of task offloading and resource allocation of multi-terminal competitive computing services in edge networks,a joint management and control optimization strategy of terminal offloading selection and communication resource arrangement is proposed.In this paper,the optimization objective is to maximize the comprehensive computing utility of multi-terminal competitive computing services.By comprehensively considering terminal access number,terminal equipment capacity,terminal demand priority,wireless channel status and computing server load,the model and control mechanism of multi-terminal competitive computing service system based on SINET and EC are proposed.The joint optimization model of terminal offloading selection and communication resource arrangement is constructed.An algorithm framework combining multiple methods is proposed.The neural network and threshold judgment method are used to estimate the priority of multi-terminal task offloading.The one-dimensional optimal search method is used to arrange the communication resource blocks.Experience playback and gradient descent are used to construct the updating mechanism of the neural network model.Distributed sampling training method is used to implement an efficient neural network training model.A vertical federated learning based flexible offloading orchestration(VFL-FOO)algorithm.The parameter tuning and performance evaluation of the algorithm are carried out through simulation experiments.The simulation results show that the algorithm has good convergence and low complexity,and the proposed strategy effectively improves the multi-terminal comprehensive computing capability under different terminal numbers and dynamic environment conditions.(4)Aiming at the problem of task offloading and resource allocation of edge convergence computing service in edge network,this paper proposes an optimization strategy for the joint management and control of robust traffic classification and resource perception forwarding.In this paper,the optimization objective is to minimize the comprehensive processing cost of edge convergence computing service.By comprehensively considering such factors as request transaction volume per unit time,arrival speed of service data volume,computing load per unit bit,type of service traffic,server computing resources and communication resources of transmission path,an edge convergence computing service system model and control architecture based on SINET and EC are proposed.A joint optimization model of robust traffic classification and resource perception forwarding is constructed.An algorithm framework including abnormal demand detection,traffic feature classification and task offloading forwarding is designed.The genetic evolutionary algorithm based rapid classification and forwarding(GE-RCF)algorithm is implemented.The parameter tuning and performance evaluation are carried out through simulation experiments.The simulation results show that the proposed strategy can effectively improve the efficiency of traffic classification and reduce the comprehensive processing overhead of the aggregation computing service under the conditions of dynamic traffic changes,different edge server performance and different propagation path bandwidth.