I am fighting this battle right now because our history has been to throw and extra cpu or memory at a VM when there is any performance problem. The problem is that the active memory for most of my machines is like 2-10%, but the OS says it is taking up almost all of the memory that I have given them. The battle is explaining that active memory is the real amount of memory that the VM is using and if you take the average for that number and then give it an extra cushion, there should be no performance loss.
One example I have right now is for an exchange server. The server has 24GB of memory assigned to it, but vmware shows the active memory is only around 2GB. I should be able to give this server 3-4GB of memory and it should be happy. I don't know why the memory explinations in vmware have to be so difficult. Fixing the descriptions for "consumed", overhead consumption, private, and active" would be of great benefit.
Thanks,
Matt