IMO generally there are few actually useful specific "best practices" for a specific application running on a general purpose hypervisor such as ESXi.
Sure there are some so called "best practices" for other popular applications or other virtualization solutions, but in the end they just boil down to common sense and simply applying existing recommendations that go for physical deployments as well: Size your VM accordingly, make sure the storage can provide the IOPS, tweak the application as you'd usually do, test and validate.
In the end you should take "best practices", like the balloon driver thing with a grain of salt and validate them against your own infrastructure to make an informed decision. I've seen lots of "best practices" that were blindly applied even though they made no sense, with admins insisting that it was absolutely necessary "I don't know why but because it's a best practice, that's all that counts".
Just my 2円 on that topic.
--
http://alpacapowered.wordpress.com
