Research And Improvement Of The Load Blalancing Technology Based On Nginx Server Cluster

With the rapid development of the economy and the science technology of our country, network has spread to most cities and families in our country.In the cities with middle or above income level, almost everybody has a computer or a smartphone People use network anywhere at anytime which makes a large number of network requests, and this put forward higher requirements for the server performance provided by the Internet service providers. But simply upgrade the server may lead to service disruption and will consume a large amount of manpower and material resources. Therefore, the server cluster technology and the load balancing technology can effectively reduce the stress cost by a large number of web concurrent requests and improve the performance of the backend servers. However, some load balancing strategies work not that well, they may cause an unbalanced phenomenon which is distributing high concurrent web requests to a busy server while other idle ones receive nothing.In this thesis, the author sets up a server cluster based on Nginx according to the actual situation, and configures the Nginx reverse proxy function modules to realize the load balancing function of this server cluster system.Aiming at the problems mentioned above, this thesis put out a dynamic self-adaptive load balancing algorithm applied in the Nginx servers cluster. This algorithm realized the dynamic calculation of the servers’weights according to the load status on the servers. It’s a good solution to improve the performance of the server cluster.In this thesis, the author collects the response time of the servers in the cluster which recorded on the reverse proxy server in a period of recent time. By making some mathematical analysis of those response times, the author calculates the average response time of the last response time of all servers and uses this average value as the first judgement condition. If the last response time of the server is lower than the average response time, it means that this server is currently in a lighter load condition, on the other hand, refers to the server is in a heavy load condition. On the other way, the author uses the server response time average variance as the stablility judgement factor of the server’s load situation.The author uses this factor in the dynamic calculation of server’s weight, and in this way we realizes the adaptive adjustment of server weights in the load balancing strategy according to the backend server load status. In this way, this algorithm realized to make full use of the resources of the backend servers in the cluster.The research in this thesis has greate significance in theory and practical application.