Scalable Framework And Task Offloading Strategy For Edge Computing

With the development of edge computing and Io T,the number of mobile terminals and Io T devices has proliferated,generating a large amount of data at the edge side of the network.However,the traditional cloud computing model is unable to meet the latency and storage requirements of intensive task computing.To meet the new demands of edge-side tasks,there is an urgent need to push intensive tasks such as artificial intelligence to the edge side of the network for execution.In addition,the current Industrial Internet architecture uses a less efficient data model building scheme and a relatively static,coarse-grained heterogeneous resource orchestration scheme,which have many challenges in terms of service scaling and heterogeneous resource access.This paper addresses two aspects of task offloading strategies and industrial Internet edge computing architectures,with the following main research elements:(1)A delay-and energy-aware Heuristic Task Multilateral Co-Offloading(HTMC)algorithm is proposed for the problem of wasting resources in the idle state of edge servers when multi-task and multi-edge server offloading exist.First,after grouping the populations by using the high and low fitness values in the particle swarm algorithm,the diversity of the populations is enriched by a particle iteration strategy based on crossvariance.Then,by analyzing the recent particle position data,a Particle Recent Pastbased Position Updating Strategy is proposed to accelerate the convergence of the algorithm by using the historical information.Simulation results show that the proposed HTMC algorithm has higher and smoother performance,smaller algorithm running overhead,and more outstanding optimization effect compared with the offloading strategy using particle swarm algorithm and genetic algorithm.(2)To design a scalable edge computing architecture with loose coupling of edge clouds for the industrial Internet domain by addressing the problems of difficult access to many heterogeneous resources and service orchestration to coordinate heterogeneous resources in the industrial Internet architecture.Firstly,a stable communication architecture based on Web Socket and backoff algorithm is designed to solve the problem of List-Watch mechanism causing abnormal occupation of control plane resources when the edge cloud network is unstable.Second,based on the middleware idea to design a heterogeneous resource access adapter,heterogeneous devices can be dynamically accessed by the heterogeneous adapter after encapsulating the driver.Finally,based on the above-mentioned edge cloud stable communication architecture,a cloud edge collaborative service orchestration architecture is designed using Kubernetes,and through the cluster management mechanism of Kubernetes,automatic scheduling and load balancing of resources can be realized to improve the fault tolerance and scalability of the system.(3)Design and implement an edge computing system for industrial vision,based on the above HTMC algorithm and scalable edge computing architecture,and validate the application in a real smart factory scenario.A real-time determination model of the working business process is constructed by integrating the new applications of human posture monitoring and target detection to monitor the factory production environment in real time.