Redundancy Strategies

Most solutions intended for production use include some type of redundancy. Redundancy strategies use multiple instances of a component to provide a single service. Redundancy is used to satisfy quality of service requirements. For example, redundancy is used to increase throughput in order to satisfy performance requirements, or to avoid a single point of failure to in order satisfy reliability requirements.

Three strategies are available for using redundant instances of Java ES components: load balancing, clustering with Sun Cluster software, and Directory Server multi-master replication. The recommended installation and configuration procedure for each of these strategies is outlined briefly in the following paragraphs:

• Load balancing can be implemented either in hardware or software. Load balancing is best set up by installing and configuring one instance of the load-balanced component, and then testing that the service provided by the first instance is available through the load balancer. After verifying that the service is available, you install and configure the additional instances of the component required by your deployment architecture. This phased approach to installing and configuring facilitates troubleshooting configuration problems.

• Clustering is implemented in several steps. The first step is to install the Sun Cluster software and establish and configure the cluster. The next step is to install the components that run in the cluster. For example, the first step towards implementing the cluster shown in Figure 2–1 is installing Sun Cluster software on computers mscs01 and mscs02, and establishing and configuring the cluster. The second step is installing and configuring Messaging Server and Calendar Server. The third, and final, step is installing and configuring the Sun Cluster Agents for Messaging Server and Calendar Server. When the Sun Cluster Agents are configured, the cluster nodes recognize the Messaging Server and Calendar Server instances.

• Directory Server multimaster replication is also implemented in several steps. The first step is installing, configuring, and verifying all of the Directory Server instances. The second step is shutting down all but one of the Directory Server instances. The third step is installing and configuring the other components in the solution. Any changes to the schema or directory structure are made to the single running Directory Server instance. The final step, after all component instances in the solution are installed, configured, and verified, is restarting the other instances of Directory Server and using the replication feature to configure synchronization and failover. This copies the modified and updated directory data to all of the Directory Server instances.

When your deployment architecture uses any of these redundancy strategies, you must develop a plan for installing multiple instances of a component and configuring the instances to operate as a single service.