There are two thresholds at the datastore or datastore cluster level:

Because fast and thin provisioned VMs and their snapshots can grow, make sure there is enough reserve to prevent an out-of-space condition on the datastore;