Research And Application Of Server Cluster Load Balancing Algorithm In Business System

With the popularity of online shopping and the rapid development of electronic commerce,after entering the 21 st century,the number of Chinese online users continues to grow.According to statistics,China's online shoppers scale has reached 500 million people.However,with the increasing of users,the visits of electricity system presents the explosive growth.Therefore puts forward higher requirements on the network server.When there is a large amount of concurrency in the system per unit time,it is likely to result in a long response time,poor stability and even a system crash.According to the above problems to carry out research and analyzes that the most effective technical means to solve the problem is to build a cluster system.So that the application servers in the cluster can jointly handle the request tasks to meet the user's access requirements in a short time.However,the core problem of the cluster system is how to make each server node reasonably allocate the request task,prevent the load imbalance and achieve the efficient utilization of resources.The quality of the load balance scheduling strategy directly affects the performance of the whole cluster.Analyzed the weighted polling algorithm,through its research work principle and the code structure,summed up the limitations of the proposed algorithm.The principle of dynamic feedback,the weighted polling algorithm is optimized.Load balancing algorithm based on dynamic feedback mechanism,the algorithm considering the cluster system in the actual operation situation.According to the load of each node of the weight of each server to make dynamic changes,the right to modify the load of each server node in the scheduler values in order to achieve the goal of dynamic load balancing,and set up based on Nginx and Tomcat server cluster system.By analyzing the built-in data structure,core module and HTTP request processing process of Nginx,the improved algorithm was specifically implemented,including the process of collecting load information of each server and modifying weight according to load information data.Finally,the cluster system was built and the apache-jmeter pressure measuring tool was used to test the performance of the original weighted polling algorithm and the optimized load balancing algorithm.The test data before and after the improvement of the load balancing algorithm are obtained through experiments.After comparison and analysis,it is concluded that when the number of concurrent connections of the improved load balancing algorithm reaches more than 1500,the overall performance of the server cluster is better than the original weighted polling algorithm,which is suitable for the system with relatively high concurrency.This research result is of great significance in the future practical application.Figure 30;Table 13;Reference 49...