For a failover cluster, the number of elements that must be online for a given cluster to continue running. The elements relevant in this context are nodes or, in some cases, a witness disk or witness file share.