Research And Optimization For High Concurrency Cluster Load Balancing Strategy Based On Nginx

With the rapid popularization of the Internet,"The Internet for All" has become a general trend.The rapid development of Internet technology has brought an explosive growth in Web access demand,which has put forward higher requirements for the performance of servers.Cluster technology can effectively cope with high concurrency scenarios,but how to reasonably allocate requests to the nodes in the cluster remains to be discussed.Therefore,load balancing technology has emerged.As the server with the highest market share,Nginx has strong concurrent processing ability and high reliability,its built-in load balancing strategies have great research significance.To meet the demands of improving cluster data processing efficiency in different concurrent scenarios,solve the problems of Nginx load balancing strategies,improve the request processing ability of server cluster,this paper conducts the following research:(1)Through in-depth study of Nginx architecture,reverse proxy and modular structure,and performance comparison between Nginx built-in load balancing strategies and common dynamic load balancing strategies,it is analyzed that current load balancing strategies have the problem of unreasonable design of server performance evaluation model.Based on this research,this paper takes the resource types in the Internet and user experience as the basis,selects CPU utilization,memory utilization,bandwidth and disk IO as evaluation indicators,designs a new server performance evaluation model through analytic hierarchy process.(2)In response to the obvious performance degradation of Nginx’s built-in weight round robin algorithm in high concurrency scenarios,this paper takes the weight calculation scheme of node real-time load as the basis,uses the server performance evaluation model designed in this paper to measure the remaining resources of nodes to calculate the weights.After studying the congestion control technology in TCP network,the AIMD idea was introduced to adjust the weights of nodes,thus generating AHP-DWCC load balancing algorithm.(3)Aiming at the issue of AHP-DWCC algorithm not considering the consumption of the requests’ resources,this paper takes the weight calculation scheme of balancing nodes’ load and request resources as the basis,proposes a load balancing strategy based on enhanced grasshopper optimization algorithm to improve the accuracy of weight calculation.This paper establishes a resource balance model by considering real-time load of nodes and requests’ resources,introduces grasshopper optimization algorithm,improves the convergence speed and optimization accuracy of the algorithm by combining the ideas of reverse learning,adaptive step size adjustment and segmented search.The enhanced algorithm is applied to calculate nodes’ weights in the load balancing strategy,thus generating EHGOA load balancing algorithm.(4)Aiming at the problems of long response time and easy loss of requests in current load balancing strategies in high concurrency scenarios,this paper designs AHP-DWCC load balancing algorithm and EHGOA load balancing algorithm based on two node weight calculation schemes,and the concurrent access threshold is set through experiments to verify the performance of the two load balancing algorithms proposed in this paper in different concurrency scenarios.When concurrent access volume is below the threshold,AHP-DWCC load balancing algorithm is used to calculate nodes’ weights,and the load balancing threshold is generated by measuring cluster’s busyness degree and load dispersion as a basis for whether nodes’ weights need to be modified;When concurrent access volume exceeds the threshold,EHGOA load balancing algorithm is used to calculate the weights of nodes.By introducing two dynamic load balancing algorithms,cluster load balancing threshold and concurrent access threshold,the advantages of the two algorithms are maximized.This paper is based on the operating environment of the student engineering system,using OPNET simulation platform and Siege tool to simulate high concurrency scenarios.The paper compares the performance of AHP-DWCC algorithm,EHGOA load balancing algorithm with weight round robin algorithm,probability preference algorithm,GA load balancing algorithm and GOA load balancing algorithm to verfify the effectiveness of the algorithms proposed in this paper.At the same time,combined with the engineering system project,the performance of the algorithm proposed in this paper is compared with the least connection algorithm to verify the practicality of the algorithm proposed in this paper in the student engineering system project.