How does vSphere High Availability calculate the memory slot size of a virtual machine?

vSphere HA calculates a memory slot size by taking the largest memory reservation among all VMs in the cluster and adding the memory overhead of that same largest VM. This ensures every slot is large enough to accommodate any VM that might need to be restarted on a failover host. The overhead accounts for VMkernel memory used for virtualization of the VM.

Subtracting overhead from the reservation would undersize the slot, meaning a slot might not hold a VM that requires both its reservation and overhead memory.

Using the overhead of the smallest VM rather than the largest would undersize slots for VMs with higher overhead, leading to incorrect admission control calculations.

Subtracting the smallest VM's overhead produces the smallest possible slot value, which would be inaccurate and allow more VMs than can actually be restarted during a failure.