• When calculating the percentage reserved for the admission control policy, the service provider must consider the cluster size and SLAs provided to the consumers on service uptime. This typically takes into account both unplanned outages, caused by hardware failure or human error, as well as planned maintenance. The larger the cluster capacity, the larger the requirement for spare capacity, because the typical mean time to failure of a hardware component within a single cluster increases. The following table provides guidance for the percentage based admission control policy for the most commonly configured building block cluster sizes such as 8, 16, 24 nodes, and so on. For business critical production systems, VMware recommends that service providers provide a minimum of 1:8 to 1:10 reserved resource capacity for admission control. This means for a 24-node compute cluster, three nodes would provide the reserved capacity to tolerate multiple host failures, or a host failure during a maintenance period where one or more hosts were already unavailable for use (for instance, during an orchestrated patching cycle).