Study On Session Maintenance In Load Balancing Cluster

With the rapid development of the Internet, the growth of users and applicationson the Web server performance and reliability of a higher demand, cluster loadbalancing technology to become one of the effective way to solve the performancebottleneck, it with scalability, cost-effective, high availability, has been widelyused.However, due to the Web technology based on HTTP protocol is stateless connectionprotocols, Web dynamic Web site will use session to record the interaction of theclient and server. When in a load balancing cluster environment, usually use sessionmaintenance technology in order to achieve the consistency of a session. At this time,the load balancing algorithms and the session remains method have an effect on loadbalancing cluster performance and load balancing.From both theoretical and practical aspects, according to the principle of loadbalancing technology and the HTTP protocol, common load balancing algorithms andsession maintain technological research; analysis of the Web server and client sessionhandling, as well as different session remains in session differences in the way ofdata storage management. An adaptive session sharing mode will be introduced toovercome the shortcomings of the existing models. On the one hand by an adaptivescheduling algorithm, the load-balancing network traffic can be dynamically adjustedbased on the node working performance and current load; the other hand, to modifythe session management Web application server, session data across serversstatistical analysis requires mapping based on the NoSQL database view of the documentmodel using JSON data structures passed and parsing the session data value; whileusing asynchronous read and write model, with remote backup to local methods to solve the problem of network access session data overhead, they also meet high availabilityrequirements. Finally, following Nginx’s load scheduler and PHP session model forpractical research, comparative performance testing, and the demonstration of thenew model can effectively improve the performance of the cluster system, reducingsystem latency time and improve system utilization and throughput, and to meet theapplication itself requirements.