Research And Application Of Load Forecasting Based On BP Neural Network

With the rapid development of IT industry,Internet technology makes people’s life more convenient,However,due to the increase in the number of Internet users,the number of concurrent business visits shows an explosive growth trend,which brings great pressure to the server.In order to achieve the high availability of the service system,server cluster technology came into being.In order to improve the overall performance of the cluster and give full play to the role of each node in the cluster,it is necessary to select efficient load balancing algorithm and appropriate load balancing scheduling strategy to reasonably distribute user requests,avoid load imbalance to the greatest extent,effectively improve the utilization of server resources and optimize the performance of the cluster system.By studying and analyzing the working principle of the traditional load balancing scheduling strategy,this paper summarizes the limitations of the traditional scheduling strategy,and in view of these limitations,proposes a dynamic load balancing scheduling strategy based on BP neural network,which mainly completes the following work:1.The commonly used load balancing scheduling strategies are studied and analyzed.Aiming at the problems of insufficient collection of server load evaluation indicators,inaccurate weight setting and unable to feed back server performance in real time in traditional algorithms,a dynamic load balancing scheduling strategy is proposed.Firstly,select the appropriate server load evaluation parameters through comparative analysis,select an optimized weighting function to measure the server computing power based on the study of the existing comprehensive load parameter evaluation model,and design a BP neural network model,which takes the server comprehensive load data of a time series obtained after collection and calculation as the input,The BP neural network is trained to predict the server load of the next time series.On this basis,the load balancer is realized based on nginx.When the client initiates a request,the load balancing server will select the node with the lowest prediction value to execute the request.2.Design and implement the online first aid training platform to verify the superiority of the request distribution strategy based on BP neural network in the high concurrency environment of real web system.The platform includes the administrator side and the student side.The core functions of the administrator side include first aid knowledge management,user management and training management.The core functions of the student side include user registration and learning,online assessment,obtaining certificates,etc.3.The load scheduling strategy proposed in this paper is tested and compared with other algorithms.The performance of other algorithms and the request distribution strategy proposed in this paper are tested by the autobench performance test tool,and each load scheduling strategy is evaluated by the request response time and the number of real connections.The experimental results show that the dynamic load balancing scheduling strategy based on BP neural network has good robustness in high concurrency scenarios.