Research On Edge Network Load Distribution Algorithm Based On Application Prediction

With the rapid development of smart devices,more and more network terminals are connected to the Internet.A large amount of network data puts huge pressure on the load of the core network.In this context,mobile edge computing came into being.By deploying edge servers at the edge of the network,the real-time processing capability near the device is strengthened.However,when the access equipment of the edge server is densely deployed,how the edge server with limited resources allocates its own resources has become a key factor affecting the end user experience.This article proposes a new solution.First,a forecasting mechanism is introduced in this article.By analyzing and processing the historical resource application records of edge devices,it is converted into a time series,and an algorithm model based on long-short-term memory neural network Make predictions and predict the tasks that the access device may use in the future according to the task sequence that has already arrived.Secondly,according to the prediction results of the model,before the real task arrives,it is distributed to different edge servers in advance to connect and load the services in the virtual machine.In this paper,the problem of resource load allocation is defined as a global optimization problem that maximizes system returns.This problem has been proved to be an NP-hard problem.In this paper,the prediction information of historical execution data is used to resolve it into two sub-problems,which are the node allocation problem and the resource allocation problem within the node Solve within.Finally,through comparative experiments,it is proved that the algorithm proposed in this paper can effectively reduce the response delay of the device and the average completion time of the task sequence through the mechanism of early loading,and the distribution result is close to the theoretical optimal.Compared with the traditional algorithm,the response time has a performance improvement of about 45%-55%,and the average completion time has a performance improvement of about 5%-10%.The research results of this article have high reference value and practical value to this field.