It really depends on what runs on these VMs, what you really need is an estimate of the peak and average CPU, Memory, I/O and Network usage for each VM and then apply a factor of overcommitment. In our environment, we try and stick to 20 VMs with 25% over commitment and that has worked well so far. In your case, clustering 2 or more ESX to host 60 VMs and then let DRS handle the load balance and HA take care of failures should work. Obviously, the user experience in case of failure (i.e. 60 VMs on 1 ESX) will not be as good as 30 on 1, so you need to set user expectations accordingly
Mehul
PS: if you find responses helpful, please consider awarding points!