We use a very simple load-balacing scheme, much simpler than what DRS does (and this does not require a DRS license). When doing a failover, HA will pick the host that is the most unreserved capacity as a percentage of that host's total capacity (this is a simple reflection of how loaded it is). Since a host has both cpu and memory resources, we simply use the average of the 2 numbers. Here's a simple example:
Host A:
Total memory=10GB, Unreserverd memory=5GB => memory available=50%
Total cpu=10GHz, Unreserved cpu=8GHz => cpu available=80%
resource available = (50 + 80)/2 = 65%
Host B:
Total memory=20GB, Unreserved memory=10GB => memory available=50%
Total cpu=10GHz, Unreserved cpu=5GHz => cpu available=50%
resources available = (50 + 50)/2 = 50%
So in this case the failover is done to Host A.
