"It would be nice if there was a memory range that could be provided to each sever and VMware can use performance counters to establish peak performance or balanced overall performance of the VMs and dynamically distribute within given ranges."
By default, DRS usese active memory to balance resource rather than consumed, which is what youre looking at there. If you also have vROPS you can make use of proactive DRS... it looks at VM utilisation over time and preemptively moves VMs when it knows theyre likely to require additional resource etc
Unfortunately its just the way Windows works... if you have Linux VMs you will probably see that their consumed isnt as high