Research On Task Optimization Scheduling Algorithm In Edge-cloud Industrial Internet

With the rapid development of Industrial Internet,a large amount of IoTs(Internet of things)devices are introduced to monitor and optimize the process of industrial production.The traditional Cloud Computing has been no longer suitable for the massive data processing and analysis due to the rapid increase of the production data.In the Cloud Computing,the upload of massive data will not only cause great pressure on communication lines,but also challenge the computing power of cloud centers.And the cloud centers are often geographically far from the production site,so it need additional time for the system to upload the data to cloud centers for processing or download results,which is unacceptable in some industrial production scene.As a new computing model,Edge Computing extends cloud computing to the edge of the network which can make full use of the computing resources in the industrial field.The Edge Computing could not only enhance the processing capacity of the system,but also reduce the delay of the system.Therefore,the cloud-edge architecture of industrial Internet has attracted wide attention,and the issues of cloud-edge computing task scheduling have attracted extensive attention too.In this dissertation,the structure and characteristics of edge-cloud industrial Internet system(cloud-edge system)are analyzed systematically.Based on the distribution and characteristics of edge nodes,the model of edge-cloud system is established.And the research status of cloud computing and edge computing task scheduling is introduced in this dissertation.In order to reduce the task completion time,the static task scheduling strategy and dynamic task scheduling strategy is proposed.The main research contents and innovative results of this dissertation are as follows.(1)Considering the problem of frequent data file transmission during edge nodes in the traditional task dynamic scheduling strategy and the need for real-time autonomy of some equipment in the production process,a static task scheduling strategy for cloudedge system based on queuing theory is proposed.The queuing theory is introduced in the strategy to calculate the expected completion time of the task which is assigned to the node in a static scheduling way.A mathematical model is established to minimize the maximum node’s task completion time.Then,a genetic algorithm based on multielite retention strategy is proposed to solve the model and obtain the optimal scheduling scheme.The experimental results show that the static scheduling strategy in this dissertation can reasonably assigned the task to the appropriate processing node,which can avoid the frequent transmission of data files between edge nodes and effectively reduce system delay.The static scheduling scheme proposed in this dissertation provides a new solution for the task scheduling problem in edge-cloud system.(2)In order to solve the scheduling problem of tasks which cannot be assigned in a static way and overcome the disadvantages of the traditional dynamic scheduling algorithm,a cloud-edge architecture dynamic task scheduling strategy based on deep learning is proposed in this dissertation.In this strategy,the deep neural network(DNN)which can be trained offline is introduced to construct the dynamic scheduler for the optimization of task scheduling.In order to get better training effect,a mathematical model of cloud-edge architecture dynamic task scheduling is established.Then,the genetic algorithm is introduced to solve the model and obtain the optimal scheduling scheme.So,the optimized training data is obtained by combing the size of the task data file with the optimal scheduling scheme.The experimental results show that the dynamic scheduling algorithm based on deep learning can get the optimal scheduling scheme quickly and effectively reduce the task completion time and system delay.