Research On Clustering Performance Of Web Application Servers

With the development of Internet,the Application Servers provide services for more and more users. In such case, the single server seems to be limited and may become the single point of failure, even though the capability of the server is quite powerful. At present, J2EE has been one of the most popular technologies used in enterprise applications for the design and the development, and J2EE clustering can provide enterprise applications with scalable and high available services. Therefore, load balancing, failover and J2EE application deployment pattern become the hot topics of increasing the clustering performance for J2EE application servers.At first,J2EE distributed multitiered application model and core technologies are introduced in this paper, and then J2EE clustering key technologies and implementations in different tiers are summarized. Besides, clustering performance is quantitatively researched and J2EE applications'architecture is studied based on JBoss, which is an open source Web application server. Considering the high cost of global session replication, a 2×n topology architecture is presented. The experiment result shows that the clustering topology can reduce the HttpSession replication cost and improve iii the system performance. Performance test and theory study prove that the performance of EJB is weak, furthermore, the performance and scalability of J2EE applications are affected by EJB technology and distributed structure. At last, by combining the average response time with weight-based Connection, a dynamic load balance algorithm with mixed factor is put forward in Web tier. The experiment result shows that this algorithm is useful.