The Research And Implementation Of Cluster Middleware For RMI Applications

The recent explosive growth of the Internet has brought with it a need to support overload and rapid load-expansion over the network. That a set of low-cost independent computers interconnected by a network can achieve or even surpass the ability of the costly Super Computer is regarded as the mainstream in the current parallel processing field.Java is a popular network development language. It is powerful on developing distributed network application. Java Remote Method Invocation is one of the most important solutions for developing network application that is programmed by Java. Now there are a lot of applications based on RM1, when the data we should processed is very large, one server will fail .So we should develop a cluster service for RMI applications.Middleware based on computer hardware and operating system, is system software on which we can develop and run application. This paper analyzed the characters of cluster middleware base on RMI, researched and implemented a load balance, scalable, manageable and transparent cluster service for RMI applications.After introduced some important technology such as JMX, Jgroup, this paper gives the mainframe of cluster middleware. The cluster middleware uses JMX framework and Jgroup for communicating. It provides some service such as Invocation and Event Dispatch Service, replica and recovery service, In-Memory Data Sharing service, HA-RMI service and application farming service.In the second part, this paper introduced replica and recovery service, In-Memory Data Sharing service, HA-RMI service. In-Memory Data Sharing service is composed of Node-related data Sharing and Global Data Sharing. It provides the local copy of data In-memory, and we can visit the shared data in cluster as visiting local ones. Replica Service manages that one member of the cluster saves its copies on the other members. When one Node of the cluster crashes, the Recovery Service can copy its running Environment to other Nodes. HA-RMI charges of how to obtain a Stub, how to choose the load balance policy and dispatching the requests. Last, we modify the Event Engine, and develop it on the cluster middleware.