High availability is achieved by removing the conditions under which one failure interrupts service. A single power supply, controller, switch, host bus adapter or cluster node is a single point of failure (SPOF), and any one of them can take the application offline regardless of how well the rest of the stack performs.

Redundant components address the hardware layer. Dual power feeds, mirrored or RAID-protected drives, paired controllers and dual fabrics allow the surviving element to carry the load when its partner fails. Clustering software addresses the logical layer. It monitors node health, fences a failed member and restarts or fails over the workload on another node, usually without operator action. The two measures are complementary, because hardware redundancy without cluster logic leaves the application stranded on a dead host, while clustering without redundant paths simply moves the outage.

The other options misstate the purpose. Redundancy does not raise the number of users an application can serve; that is a scaling concern. Compatibility with virtualization is governed by support matrices rather than by redundancy. Redundant infrastructure generally raises, not lowers, cost, and that cost is justified by the avoided downtime. Eliminating single points of failure is the defining objective of an availability design.

Study Guide reference: HPE Storage Fundamentals (HPE0-J81), data protection theory; availability, redundancy and fault tolerance design principles.