I have found that when virtual memory is assigned to a vm the system assigns half of it to physical memory and half to the swap file memory. You also have the option of reserving physical memory for each vm. This means that there must be an amount of physical memory equal to the reservation free in order to power on the vm. This does not mean that the vm will use all of the memory however, if you assign 1GB but the vm only uses 512 the system can reallocate the other 512 somewhere else.
Hope this helps.
The "official" def from VMware:
"Guest memory usage is based on the amount of memory that is actively being used by a process. If you would like to test this, simply run a batch loop script inside the Virtual Machine and you will notice that the Guest memory usage will increase."
I don't see the point; the "host memory usage" seems more beneficial to me. But that's just me
1 person found this helpful
I like the Guest Memory Usage better. Take an example of a SQL/Exchange server (which will chew up all the RAM that it can whether it needs it or not) with 2GB assigned to the VM. the Host Memory usage will show almost all 2GB in use, while the Guest Memory Usage will show you how much of that 2GB is truly needed and actively used and not just "held" by the application.
I have a 2003 Exchange server with 4 mailboxes that shows 1.9GB Host memory usage but the Guest usage only shows 5% (which is truly how much memory is needed). I just haven't migrated the other 200 mailboxes yet, so I know the guest memory usage will go up considerably.
Now I'm not saying to drop your RAM for each VM so the guest memory usage is 80%, but there is no need in wasting RAM on a VM just because certain applications will take whatever you give it and then Task Manager shows you only have a few hundred MBs left. Just watch your performance charts and you'll quickly see what VMs are truly using the resources and which ones aren't.
Ahhhh, ok....now we're starting to make sense. I was trying to understand the application of the "guest memory usage" stat, and that definitely applies. I guess I have yet to see that scenario myself, but thank you.
Then you're looking at the host server total memory utilization. That would be the physical RAM in use, which would encompass all virtual machines running, the virtualization overhead, the service console, etc. If you're consistently spiking that and generating warnings, you either need to decrease the memory allocation for some virtual machines or add more physical memory to your server.