The Improvement And Implementation Of Load Balancing Strategy For WEB Server Based On Nginx

The requirements of the network server are higher with the rapid increase of the equipments and users that access to the Internet in modern society.Users have increasingly high requirements for the performance such as concurrency value,resource utilization and response time and so on,and adopting Nginx server cluster system is frequently-used solution with better effect,in which one core problem is the load balance,which affects the system performance directly.The present study improves a dynamic load balancing based on the Nginx original weighted round-robin through the analysis of Nginx system source code,data structure,main module,load balancing,architecture as well as the principle of the main process and work process.The load data collected in this algorithm mainly include node CPU performance,disk I/O performance,memory performance,network bandwidth performance,which can be converted to the corresponding weights by using dynamic weighted round-robin and modify the weight value in Nginx weighted round-robin to achieve the purpose of dynamic load balancing.The experimental test environment constructs Nginx and Tomcat server cluster systems on the bases of Linux system.Through the comparison of Nginx weighted round-robin as well as the fair,a third-party load balancing with improved dynamic weighted round-robin in this paper by using performance testing tool—Httperf and Autobench,we can find that the peak value that dynamic weighted round-robin responses per second and the average of load in the saturation state increases by 17.4%and 20.3%respectively than static weighted round-robin,and increases by 6.9%and 8.3%than fair,the load balancing algorithm of third-party.Dynamic weighted round-robin has shorter response time and smaller jitter of response per second than the other two algorithms.Experimental results verify that the improved dynamic weighted round-robin is superior to Nginx built-in load algorithm and fair,the third party algorithm.Compared with the two kinds of load balancing algorithm above,the dynamic weighted round-robin in the present study has the following characteristics:(1)Adopting node comprehensive performance to evaluate,collect related node performance data and calculate the value of node is more effective than the way that personal experience gives assignment to the node weights.(2)In terms of load determination,this paper sets up the threshold value weight.If the variance of cluster comprehensive load of each node exceeds the threshold T,it shows that cluster load imbalance needs to modify the node weights.(3)Through the analytic hierarchy process of multiple attribute decision making to calculate the weight coefficient K of evaluation index.