Host uses all the ram assing on VM's

kind

Hi,
I wanna know if this is normal or its a config issue.
I have my host with a memory usage equivalent to all the ram that the VM's has, but, the VM's aint consuming that much of memory.
Both VM's are running Windows Server 2016 with all the VirtIO drivers correctly installed and services running (qemu, balloon)
Host Usage:

VM's Usage (both VM's are equals):

VM's HW config (both VM's are equals):

Forgot this:

Any tip on what to do? Thanks!
PD: Promise is my last support topic

MaxKVM

This looks completely normal to me, though I am not an expert on proxmox. Do you have fixed or dynamic memory allocation enabled?

In Proxmox, we can set fixed or dynamic memory for a VM. Automatic range is also known as memory ballooning. For the fixed option, all memory is allocated at once. In the dynamic option, memory is allocated based on the VM, within a preset range. Automatic memory allocation works great for Linux-based guest VMs. But for Windows VMs, memory ballooning consumes a higher amount of CPU resources, causing the VM to slow down. So for windows VMs, it is best to use fixed memory whenever possible.

seriesn

https://www.linuxatemyram.com/

kind

@MaxKVM said:
This looks completely normal to me, though I am not an expert on proxmox. Do you have fixed or dynamic memory allocation enabled?

In Proxmox, we can set fixed or dynamic memory for a VM. Automatic range is also known as memory ballooning. For the fixed option, all memory is allocated at once. In the dynamic option, memory is allocated based on the VM, within a preset range. Automatic memory allocation works great for Linux-based guest VMs. But for Windows VMs, memory ballooning consumes a higher amount of CPU resources, causing the VM to slow down. So for windows VMs, it is best to use fixed memory whenever possible.

I have dynamic memory, and all the drivers required to work correctly installed. Thats why its surprise me that the memory its still allocated.

Not_Oles

my last support topic

Nope! Don't believe it!

When you want to run a program the computer must load the program from the disk into the RAM. Reading from the disk is slow.

Now you want to run a second program. The computer must read it into RAM from the disk, which is slow.

Now you want to run a third program. This also must be read slowly from the disk.

We can see that the RAM use is continually increasing.

Now you want to run the second program a second time.

If the second program still is in RAM, there is no need to read it slowly from the disk. Instead, because the program still is in RAM, it can launch practically instantly.

The whole system works faster if the kernel makes intelligent use of the RAM by keeping as many likely-to-be-used again soon programs in the RAM as possible while still leaving space for yet another new program.

If the system is working correctly, then RAM looks pretty much full all the time. Except when the RAM size significantly exceeds what is needed.

An easy way to see all this is to reboot your laptop and launch Firefox or another relatively big program. It takes a moment to load. Then exit the program. Then call the program again (without rebooting). It loads much faster because it's still in RAM.

Greetings from Mexico!

Mr_Tom

I've always found the proxmox interface reports higher RAM usage than the VPS reports.

InceptionHosting

@Mr_Tom said:
I've always found the proxmox interface reports higher RAM usage than the VPS reports.

that is because there is overhead to run the VPS from the hostnode side and proxmox includes that while obviously that overhead is not visible to the guest/vps.

kind

Thanks for your responses, but I still don't understand what ballooning is for if memory is allocated at 100% anyways.

InceptionHosting

@kind said:
Thanks for your responses, but I still don't understand what ballooning is for if memory is allocated at 100% anyways.

Very simply it allocated the memory because it is free so that if it needs it it is faster, under real contention it may not behave the same way.

Basically for efficiency.

Ballooning is more for dynamic memory use without reboots.