if the two systems do not share common disks, then the standby provides a functional
backup but has no access to the databases managed by the primary.
The passive standby is generally not referred to as a cluster.The term cluster is
reserved for multiple interconnected computers that are all actively doing processing
while maintaining the image of a single system to the outside world. The term
active secondary is often used in referring to this configuration.Three classifications
of clustering can be identified: separate servers, shared nothing, and shared memory.
In one approach to clustering, each computer is a separate server with its own
disks and there are no disks shared between systems (Figure 17.10a). This arrangement
provides high performance as well as high availability. In this case, some type
of management or scheduling software is needed to assign incoming client requests
to servers so that the load is balanced and high utilization is achieved. It is desirable
to have a failover capability, which means that if a computer fails while executing an
application, another computer in the cluster can pick up and complete the application.
For this to happen, data must constantly be copied among systems so that each
system has access to the current data of the other systems.The overhead of this data
exchange ensures high availability at the cost of a performance penalty.
To reduce the communications overhead, most clusters now consist of servers
connected to common disks (Figure 17.10b). In one variation on this approach, called
shared nothing, the common disks are partitioned into volumes, and each volume is
owned by a single computer. If that computer fails, the cluster must be reconfigured
so that some other computer has ownership of the volumes of the failed computer.
It is also possible to have multiple computers share the same disks at the same
time (called the shared disk approach), so that each computer has access to all of the