Research On Design And Implementation Of Cluster Service In Middleware

As an important component of middleware platform such as J2EE application server and distributed transaction monitor, cluster service provides diverse functions, including topology management, failure monitoring and recovering, load balancing, status replication and etc., which gains high scalability, high reliability and high availability for the middleware.The thesis introduces the design and implementation of cluster service in middleware platform, adopting cluster services in J2EE application server OnceAS and distributed transaction monitor OnceTx as case study. By adopting multiple design patterns and implementation independent of the specific middleware, the cluster service improves software reusability and provides service transparently.With more complex industrial applications, cluster service should be adaptive with the different environments, and be self-configuring, self-optimizing, self-healing and self-protecting to some degree. The thesis describes adaptive topology management and load monitoring, analyses adaptive load balancing carefully, presents the general framework of load balance algorithm, and implements customized load balance algorithm. In order to enhance the performance of the entire cluster servers, the thesis examines the technical pitfalls of common load balance algorithms, presents the fuzzy load balance algorithm for optimization, and solves the potential load thrashing problem.Further, the thesis illustrates the influences on cluster service of the efficiency of network communication, and analyses the scalability of cluster service quantitatively. Additionally, the thesis discusses how to gain a better performance of cluster servers by using adaptive load balancing, and the experiment data confirms our efforts eventually.