Hardware/Operating System Level Failover (Active/Passive Setup)

The most transparent way to implement failover is to use software packages/mechanisms from operating system vendors and third parties, such as SUN, IBM®, HP or Veritas, that allow one server to become another server, taking over its network address, its disks and its workload.

There is always some time delay involved in the switch.

Using this method requires that the system have at least one spare server. The spare could either be working on something else or it could be just waiting to be activated. If full capacity must be available, the spare must sit idle or only be used for tasks such as testing or development that can be interrupted. The spare can be configured cold, where the node is switched off, or warm, where the node and as much higher-level software as possible is running to minimize takeover time.

The following graphic depicts a simple warm failover environment in which disk connections and IP addresses are taken over by the second node after a failure is detected by the cluster management software. In a real implementation, there would most likely be redundant paths to the disk and network for each adapter as well as an out-of-band heartbeat connection using serial or parallel ports.