Research On Joint Resource Management Technologies For Edge Computing In ML-Based IIoT Applications

Machine learning(ML)plays an important role in intelligent industrial Internet of Things(IIoT)applications.Neural network inference in real time requires a lot of computing resources,which poses a huge challenge to IIoT devices with limited resources.Edge computing and cloud computing can greatly improve the processing efficiency of ML tasks.However,the existing system optimization work often ignores the difference of inference accuracy of ML models with different complexity and its effect on system performance,and lacks the joint optimization of task offloading,resource allocation and model deployment.Therefore,this paper studies joint resource management technologies for edge computing in ML-Based IIoT applications.The main contributions are as follow:(1)This paper proposes a joint task offloading and resource allocation scheme based on model inference accuracy in an edge-cloud-based network architecture.This scheme aim at minimizing the long-term average system cost affected by the task offloading,computing resource allocation,and inference accuracy of the ML models.The Lyapunov optimization technique is applied to convert the long-term stochastic optimization problem into a short-term deterministic problem.An optimal algorithm based on the General Benders Decomposition(GBD)technology,and a heuristic algorithm based on proportional computing resource allocation and task offloading strategy comparison,are proposed to efficiently solve the problem,respectively.The performance of our scheme is proved by theoretical analysis,and the two algorithms are compared with existing schemes by changing seven key parameters.Simulation results demonstrate the effectiveness and superiority of our two algorithms.(2)This paper proposes a joint model deployment,task offloading and resource allocation scheme in multi-edge network.The pre-trained DNN models with different complexity are flexibly deployed on the edge servers of different base stations,and the tasks are reasonably allocated to each base station.In order to achieve efficient task processing,a joint model deployment,task offloading and resource allocation problem is proposed to minimize the total processing delay of all inference tasks,while meeting the delay and inference accuracy requirements.To solve this problem,this paper proposes a joint model deployment,task offloading and resource allocation algorithm based on GBD technology and RL(Reformulation Linearization)technique.Simulation results show that our scheme can allocate storage resources and computing resources effectively,greatly improve the system performance,and is superior to the comparison schemes in all scenarios.The research on joint resource management technologies for edge computing in ML-Based IIoT applications proposed in this paper effectively improves the ML application performance and resource utilization of IIoT,and the research results can provide theoretical reference for related work in this field.