An uninterruptible power supply protects the cluster against power failures. If the mainline power fails to one or more nodes in the cluster, the uninterruptible power supply maintains the internal power long enough for the cluster state information to be stored on the internal SCSI disk drive of each node.
A cluster must have two uninterruptible power supply units. One uninterruptible power supply must be connected to one half of the nodes that are in the cluster. The other uninterruptible power supply must be connected to the other half of the nodes. The cluster can now continue work (in degraded mode) if one uninterruptible power supply fails.
It is very important that the two nodes in the I/O group are not both connected to the same power domain. Each SAN Volume Controller of an I/O group must be connected to a different uninterruptible power supply. This configuration ensures that the cache and cluster state information is protected against the failure of the uninterruptible power supply or of the mainline power source. If possible, each uninterruptible power supply should be connected to a different power source. Otherwise, a power source failure will result in the I/O group being taken offline.
When nodes are added to the cluster, the I/O group they will join must be specified. The configuration interfaces will also check the uninterruptible power supply units and ensure that the two nodes in the I/O group are not connected to the same uninterruptible power supply units.
The figure below shows a cluster of four nodes, with two I/O groups and two uninterruptible power supply units.

| Attention: Do not connect two clusters to the same pair of uninterruptible power supplies. Both clusters will be lost in the event that a power failure occurs on both of those uninterruptible power supplies. |