Historically VMware has suggested to methods for sizing vmfs datastores - Predictive and Adaptive
- Predictive - the predictive method you will set up different datastores with different configurations and disk performance to allow you to satisfy the disk performance requirement of the VMs OS and application requirements
- Adaptive - the adaptive method you will create datastores with the same configurations creating data stores of the same size - typically I have sized them from 300 GB to 800 GB - adding VMs to a datastore until either the disk is filled up or disk performance is affected -
In terms of free space for vmkernel swap files and logs Vmware in the past has recommended leaving 20% free - in terms of snapshots snapshots should be on for long on production machines - if you do have a need to run with snapshots I would increase the free soace -
I am also moving this thread to a more appropriate forum