Win 10 in Proxmox - Performance

I suspect the qcow2 backed disk was trying to expand/grow the virtual disk as you were filling it up.
I remember hyper-v dynamic expanding disks causing issues when growing, most noticeable on databases being wirtten out.
 
Last edited:
The only difference i figured out sow far:
1612551246491.png

Just an additional Information. To install O365 took multiple hours - with 5Mb/s hdd writing speed ☹ (on the bad VM)
Transfering the testfile from "good"VM to a NAS is working with more than 120Mb/s ... what ?!?!?:eek:
 
Last edited:
Anti-Ctrl said:
The difference is, that you are using raw in the "new" VM and qcow2 Image in the "kaka" VM. Thats the difference in Performance.
Click to expand...
You are right !! I did not see this fact!
I am sorry.


Can i change this in the config ?- w.o. killing the VM ?
 
kohle said:
You are right !! I did not see this fact!
I am sorry.


Can i change this in the config ?- w.o. killing the VM ?
Click to expand...
you can run qemu-img convert -f qcow2 -O raw /path/to/image.qcow2 /path/to/image.raw at least in theory. Take a Backup before!

EDIT: You may have to fix the path in the config after that, but I´m not sure.
 
And as Dominic suggested, I would choose "SCSI" instead of "VirtIO Block" as the vdisk bus/device type. As far as I know "VirtIO Block" is kind of deprecated.

EDIT: And your CPU type is still "Default (kvm64)"? If you don't plan a cluster with different physical CPUs you should try "host".
 
Last edited:
  • Like
Reactions: Anti-Ctrl
sherminator said:
And as Dominic suggested, I would choose "SCSI" instead of "VirtIO Block" as the vdisk bus/device type. As far as I know "VirtIO Block" is kind of deprecated.

EDIT: And your CPU type is still "Default (kvm64)"? If you don't plan a cluster with different physical CPUs you should try "host".
Click to expand...
thx for advice
i am not ignoring your proposal.
Will follow up after HDD is converted to raw.
 
Update: after converting the HDD:


1612589877969.png
The Result is much better!! o_Oo_O:):)o_Oo_O

I was testing 2 cases:
Case 1: copied 10 GB from NAS to local HDD => constant 105MB/s
Edit:- After some more testing: not reproduce able (tested with multiple different VM’s)
Case 2: copied 10 GB from HDD to HDD (picture above)

Curios Question: Why is the local copied Performance fluctuating between 30 MB/s -200 MB/s
 
Last edited:
1612596831223.png
NAS faster than Folder2Folder (my assumption was F2F would be much faster)
Different results in different VM's (even the Config is the same)
hmmm.... intressting
 
Last edited:
sherminator is right. You can test that scenario easily on your own PC. Copy a file Folder to Folder on the Same HDD and you will see speeds crippling. On an SSD thats not such a big impact, but still noticable. Its just the worst case for HDDs.
 
Changing format on CLI is fine, of course. If you prefer GUI: The move disk button lets you change the output format if the target storage is of type directory.
move.png

SSD emulation might be especially interesting together with TRIM/Discard. In the background, SSD emulation sets the rotation_rate parameter. So if you have it enabled and type ps -aux you should be able to grep for vmid or rotation_rate
Code: 
ps -aux | grep rotation_rate

There are some conditions when exactly the parameter is set. If you're a little into programming then I think it should be possible to get an idea of them by looking at the if/elsif lines surrounding the 2 occurences of rotation_rate in the code without understanding everything: https://git.proxmox.com/?p=qemu-ser...5e5e9a26b783aa997669b514d03a064;hb=HEAD#l1456
 
You must log in or register to reply here.
Top Bottom