Server clusters are based on the concept of shared software configuration
data storage, making it suitable for data-centric tiers, such as file servers and data servers
because their data content is constantly changing. Server clusters typically communicate through a
broadcast or share a central repository to keep track of cluster information and cluster node
status.
Server hardware and software vendors offer vendor-specific server clustering products as their
high availability offering for these kinds of data-centric servers.
- Two or more servers share a highly available disk array for data storage, shown in the following
figure. The array incorporates redundant copies of the data, but appears as a single disk resource
to the servers, thereby avoiding the need for data replication between servers. The servers can each
have their own local disk for static storage of operating system, utilities, and other software.
Figure 1. Basic server cluster that uses RAID storage
- A common set of applications run on each server.
- Server clients see the cluster as a single virtual server.
- If one of the servers fails, the other server picks up the workload of the failed server
(failover). When the failed server is repaired and ready to run again, the workload is shifted back
over from the other server (failback). In some configurations, the repaired server simply becomes
the new backup server, and no failback is required.
- In most server clusters, only one server is actively serving clients at a time. This is called
an active-passive configuration. Some cluster server products also support another mode, called an
active-active configuration. In this mode, all the servers in the cluster can be actively sharing
part of the workload at the same time. This configuration typically requires an application designed
to partition data sets among the servers to avoid data integrity problems resulting from concurrent
updates to the same data from multiple servers.
- Sessions for some components can be replicated across cluster nodes, eliminating the need for
users to log in again after a failover:
- IBM® Content Navigator
V2.0.2 and later and Administration Console for Content Platform
Engine support session replication.
- Session replication does not apply to the Content Platform Engine server,
because the server is stateless and does not use sessions.
- Workplace and Workplace XT do not support session replication.