Research On Load Balancing Based On Session Sharing

With the rapid development of the network, the rapid increase in user traffic, the Web server performance and reliability of a greater challenge, the use of server cluster technology and load balancing technology can effectively solve a large number of problem about web concurrent access to the back-end server pressure. Load balancing technology is the core of the server cluster system, it is related to the back-end server can maintain efficient, reliable and stable working conditions or not.When client accesses to the web, Web site will use the session technology to record the interaction information between client and server-side due to the HTTP protocol is a stateless connection agreement. When you are in a load-balanced cluster environment, session persistence techniques are often used to achieve session consistency. In this case, both the load balancing algorithm and the session holding method have an impact on the performance and load balancing of the load balancing cluster.This paper first introduced the basic concept and algorithm of load balancing technology,analyzed the common-used session keeping technology, analyzed the advantages and disadvantages of the conmmon-used session keeping technology, and finally used the independent server to manage the session data so that the session data was in a globally visible way to maintain the consistency of the session, because the session server will be frequently accessed by each node and this became one of the hot spots in the system, so the session server used a redundant backup. The overall design scheme of the system was given,and the operation flow of the whole session was described from the demand analysis. Nginx server as a load balancing server, the analysis of the Nginx source code, a dynamic load balancing algorithm was used to transform the residual load and performance ratio of the node server into weights, selected the weight of the first three servers, and then randomly selected one of the selected server allocation to assign the client request, When randomly assigned, the probability of being selected was proportional to the weight of the server,thereby enhanced the performance of the cluster,improved efficiency. Finally, used the LoadRunner performance testing tool to test the original weighted polling strategy and the dynamic strategy adopted in this paper, it was proved that the dynamic strategy adopted in this paper could achieve the load balancing better.