|Bookshelf Home | Contents | Index | PDF|
A server cluster is a group of two or more servers that are configured so that if one server fails, another server can take over application processing. The servers in a cluster are called nodes. Typically, these servers store data on a common disk or disk array.
Clustering software monitors the active nodes in a server cluster. When a node fails, the clustering software manages the transition of the failed server's workload to the secondary node. Siebel Systems has certified a variety of third-party vendors to provide server clustering for the Siebel deployment. For a list of vendors and requirements, see System Requirements and Supported Platforms on Siebel SupportWeb.
When a clustered Siebel Server fails, all the applications and services on the server stop. Application users must reconnect and log in to the server that takes over. For example, if the Siebel Server that failed was hosting Siebel Communications Server, the Communications toolbar is disabled, and users must reconnect and log in to the new server.
An active-passive server cluster contains a minimum of two servers. One server actively runs applications and services. The other is idle. If the active server fails, its workload is switched to the idle server, which then takes over application processing.
Because the standby server is idle, active-passive server clusters require additional hardware without providing additional active capacity. The benefit of active-passive clusters is that after a failover, the same level of hardware resources are available for each application, thereby eliminating any performance impact on users. This is particularly important for performance-critical areas such as the database. The most common use of active-passive clusters is for database servers.
An active-active server cluster contains a minimum of two servers. Both actively run applications and services. Each may host different applications or may host instances of the same application. If one server fails, its processing load is transferred to the other.
Potential port conflicts. Some Siebel Server components, such as Siebel Connection Broker (SCBroker), Remote Synch Manager, Handheld Synch, and Siebel Gateway Name Server listen on a configurable static port. When these components run in an active-active cluster, you must plan your port usage so there is no port conflict after failover.
For example, an active-active server cluster contains two platforms, each running a Siebel Server. If one platform fails, the other will host two Siebel Servers. Siebel Servers include a number of services, such as Siebel Connection Broker, that use a dedicated port. If this port number was the same on both platforms, there will be a port conflict after failover.
Capacity planning. Active-Active clusters use all the server platforms continuously. This takes better advantage of computing resources than active-passive clusters. When doing capacity planning, make sure that clustered servers have sufficient capacity to handle a failover. Because failovers are usually infrequent and normally last only a short time, some performance degradation is often acceptable.
|Deployment Planning Guide|