In addition, I would recommend to use Percentage Based Admission Control Policy. This is to move away from the slot size calculations from "Number of Host Failures the Cluster Tolerates". As the calculations are very conservative, ie when selecting 1 host for failure, one of the hosts with 512GB will be taken out of the equation. If you are using VM reservations, then this is another thing to consider as slots will use the highest cpu and memory reservations.
With Percentage Based Policy, just make sure the percentage is equal or larger than the percentage of resources provided by the largest host. This is to ensure that all VMs residing on the this host can be restarted in case of host failure. This is well documented in Duncan Epping and Frank Dennemans Clustering Deepdive book.
You can check out this blog which clearly explains this concept Calculating Percentage of Cluster Resources Admission Control Policy for an unbalanced vSphere 5.x cluster « D…