A cluster system uses two or more server machines as a single virtual server machine to provide high level of availability.
In single-server machine operation, if the server machine or an application running on it stops, the machine or the application must be restarted. During this time, the operation stops.
A cluster system, however, uses two or more server machines. If a failure occurs on one server machine or an application running on it, the transaction can be resumed immediately by restarting the application on the other server machine. As a result, transaction downtime is reduced. The switching of a transaction from a server with a problem to another server that is operating normally is referred to as failover.
In the cluster system, the server machines are collectively called a cluster, and each server machine in the cluster is called a node.
The cluster operation is classified into one of the following types. These types are explained in "Names of Fujitsu PRIMECLUSTER."
Standby class
In the standby class, a standby node exists to take over transactions from an active node on which transactions are being performed.
1:1 standby system
The cluster consists of one active node and one standby node. The active node performs transactions while the standby node is waiting to become the active node, if required.
Figure 1.1 1:1 standby system
Mutual standby system
This is a cluster consisting of two or more nodes. Generally, two of the nodes are used as "active nodes." Each node performs its own transaction and, at the same time, is on standby to run the transaction of the other node. If either node fails, the other node takes over running the transactions on both nodes.
Figure 1.2 Mutual standby system
n:1 standby system
The cluster consists of n active nodes and one standby node. Each of the n active nodes performs separate transactions while a separate standby node stands by to take over from any failing active nodes.
Figure 1.3 n:1 standby system
Cascade topology
This consists of three or more sets of nodes. One node in the set is an active node, and the remaining nodes are standby nodes.
Figure 1.4 Cascade system
Priority transfer
This consists of three or more sets of nodes. One node in each set is an active node, one node is a standby node, and the remaining nodes in a set are stop nodes.
Figure 1.5 Priority transfer system
Scalable class
In a scalable class, the cluster is designed for performance improvement and degraded operation when a failure occurs by means of parallel processing on multiple server machines. In contrast to the standby system, nodes in this class of cluster are not divided into active and standby nodes. If one server in a cluster fails, the remaining servers in the cluster operate at a degraded level so that the transaction continues.
Figure 1.6 Scalable class
Note
The cluster operation and the names given to those that are supported with a cluster system differ from each other in different systems. For more information on this, refer to the cluster software manuals.
