A: My last FAQ talked about the four techniques ESX 4.1 uses to manage memory during periods of contention. When there isn't enough memory to meet the needs of virtual machines (VMs), the host and its hypervisor must take actions to prevent memory from ever completely being exhausted.
ESX's four stopgap techniques are explained in great detail in the VMware document "Understanding Memory Resource Management in VMware ESX 4.1." But while memory management techniques are good for preventing a memory-related crash, avoiding needing them in the first place is never a bad idea. That avoidance means being careful with the creation of VMs and always keeping enough RAM lying around. In VMware's document are another five best practices that you might also consider when it comes to managing memory in your vSphere environment:
Don't disable page sharing or the balloon driver.
Carefully specify memory limits and reservations.
Host memory size should never be larger than guest memory usage.
Use shares to adjust relative priorities when memory is overcommitted.
Set an appropriate virtual memory size.