What is right way to use mem_overcommit_ratio setting in VSphere CPI?

[Sujeet Varakhedi]

We are trying to play around with mem_overcommit_ratio a little bit so that we can over-provision our ESX clusters for use in dev/test and CI. We looked at the BOSH source code to see how it is using mem_overcommit_ratio setting and seems like the right place to use it is in the deployment manifest like this:

properties:

  vsphere:

    mem_overcommit_ratio: 0.5

We tried different settings but we always peak out at a fixed number of instances regardless of this setting. I am wondering if it is picking up the setting at all? 

Question: 

1. Is this the right place to put the setting? 

2. Has anybody tried over commiting the ESX hosts with BOSH and CF deployments?