Design And Implementation Of Intelligent Computing Offload Mechanism For Heterogeneous Edge Network Based On Delay Guarantee

Benefiting from the development of Internet of things technology,more and more kinds of devices begin to access the network?New services such as Virtual Reality(VR),Augmented Reality(AR),automatic driving,artificial intelligence emerge one after another,which puts forward the requirements of low delay and high reliability for network computing services.Although the traditional cloud computing paradigm can provide sufficient computing resources for devices,the cloud computing center is geographically far away from users,so it is fundamentally unable to solve the problem of back and forth communication delay.As an extension of cloud computing,edge computing paradigm moves the function of cloud computing closer to the edge of users,provides users with sufficient computing and caching services,avoids unnecessary propagation delay and network congestion,and has significant advantages in delay,reliability and security.However,compared with the remote cloud computing center,the edge server has limitations in computing resources.In order to adapt to the trend of multiple growth of networking devices and provide reliable computing services for user business,it is of great significance to study a computing task unloading mechanism that can reasonably allocate limited resources.To solve these problems,this paper proposes a delay guaranteed edge computing task offloading mechanism.The specific work is as follows:1)a network model of heterogeneous cloud edge collaborative computing is constructed to solve the problem of device delay calculation in different communication modes.In this paper,different delay calculation models are designed for wired and wireless devices by setting flag bits to distinguish optical fiber link and cellular link,and the effect of one model covering two kinds of devices is realized.2)Aiming at the problem of computing offload timeout in edge network,a multi-objective optimization algorithm including task success rate and delay is designed.In this paper,based on deep reinforcement learning method,the agent is inclined to make a priority decision to ensure that each task can be completed on time.Based on the above work,this paper designs and implements a simulation verification system of edge computing task unloading algorithm.The system has built-in computing offload mechanism proposed in this paper,and provides functions such as constructing edge network,importing algorithm,computing task offload simulation,etc.it provides a simulation platform for scientific researchers to verify the effect of computing offload algorithm.Through the graphical interface,users can construct the network topology with low threshold and instantiate the edge network environment,so that researchers can focus on the implementation of the algorithm and simplify the complicated steps of constructing the edge network environment.To sum up,based on deep reinforcement learning method,this paper proposes an intelligent unloading mechanism of edge computing tasks for delay guarantee according to the business requirements of delay guarantee,and designs and implements a simulation verification system of edge computing tasks unloading according to the mechanism,which provides an algorithm verification platform for the experimental simulation of edge computing.