Ha! I wish it was. Nope this is purely budget. Basically this client is stacking as many VMs onto Hosts as possible and wants to know where the line in the sand is before the host no longer has any resources to perform. Some of the performance metric for the hosts are giving back some of that info if I look at VMKernel.
The company runs a fairly large LAMP farm. Each ESXi host is running 512GB of memory and there are currently 15 ESXi. Since its all Linux, the VMs take as much host memory as it can whether the VM needs it or not. The Operations team is working on that issue so the company can cut down on requirements (and most likely put more VMs onto hosts). I need to see what the System Reservation needs to be for the Host. The reason for this, the client has been hit too many times killing the memory on the server so that the HOST itself doesnt have enough to run internal operations. That second link you provided pushes in the correct direction I believe. Each Host has about 45 VMs (+/- 3 or 4). They each run about 16GB (again +/-). This gives me about 16GBs of overhead for the Host. Now not all the server are taking 16GB and not all servers are running at the full amount they take (love Linux). However the hosts internal performance for memory "Used by VMKernel" is 4.5GBs... this math appears to be right and fits for what the host is doing. Again this is quick glance and quick math.
So, hopefully this answer better of what I am looking for. I am trying to see what the Host is using for overhead and resources need to run "itself" AND "the VMs" that are hosted on it. This will allow me to give a proper recommendation (not quick math) of what we should set for the Hosts "System Resource Allocation" under Configuration so the host wont loose the needed Memory to operate in time of contention.
Thanks,
BostonTechGuy