Purely from performance perspective.
Goal: Ensure VM's memory allocation is guaranteed with lowest latency. VM running critical workload.
Ensure cluster memory resources are not overcommitted even during N+2 outage and take into VM memory overhead capacity buffer. This will ensure memory reclamation techniques are not triggered even during outages and the VMs get their allocated memory with lowest latency
Configure Memory reservation equivalent of VM size. This is guaranteed to work.
Thinking out loud on option 1. Are there any other scenarios where this VM will have to wait for the ESXi scheduler to reclaim memory pages prior allocating them to the VM?
Hey, hope you are doing fine.
Both approaches are fine.
If you have the most critial application and any risk to it is not an option, a reservation is a good option. but those resources will not be available to other VMs in the cluster.
Ensuring no overcomitment is another good option.
IMHO both options should work better with VM's resources optimized at it's best.
Some kind of applications like oracle DB's crash when they see a hint of resource contention.
Hope this works