Since we have some active / active clusters, I configure my max memory for each instance, so there will be enough reserved memory on a node to at least allow the failed instance to restart on the other node. As a rule, for our standalone VMs, we only install one instance per VM.

-Ray