HDD queue settings for VM

[showiproute]

Hello there,

can someone tell me the "secret" where to place queue settings for a Windows VM.
Someone from Proxmox staff told me some years ago to add this to the guest config itself - but I am unable to find this thread and also Google is not showing something useful.

Background is that I am currently playing around with different zvol blocksizes to determine which one would be the best for a sequential storage VM disk.
But currently I am already having issues with throughput.

Copy paste native to the ZFS PVE system -> 1 GB/s
Copy past to WIndows VM -> 100 MB/s

 

[showiproute]

While doing my tests I stumbled accross a very weird behaviour:
My Win22k server used IPv4 & IPv6 with dedicated addresses.
Current settings:

agent: 1
bios: ovmf
boot: order=scsi0;ide2
cores: 8
cpu: host
efidisk0: local-lvm:vm-109-disk-0,efitype=4m,pre-enrolled-keys=1,size=4M
ide2: none,media=cdrom
machine: pc-q35-8.0
memory: 16384
name: WindowsServer
net0: virtio=3E:0E:15:D6:44:5A,bridge=vmbr0,queues=8,tag=10
numa: 1
onboot: 1
ostype: win11
parent: autohourly231024090523
scsi0: local-lvm:vm-109-disk-1,discard=on,size=200G,ssd=1
scsi1: Data:vm-109-disk-3,discard=on,size=2T
scsi2: Media:vm-109-disk-2,backup=0,discard=on,size=10T
scsi3: Data:vm-109-disk-0,backup=0,discard=on,size=3T
scsi4: Data:vm-109-disk-1,discard=on,size=4T
scsi5: Media:vm-109-disk-1,backup=0,discard=on,size=11T
scsi6: Data:vm-109-disk-2,backup=0,discard=on,size=500G
scsi7: Media:vm-109-disk-3,backup=0,cache=unsafe,discard=on,size=100G
scsihw: virtio-scsi-pci
smbios1: uuid=4e7d194d-448c-4585-b349-766632e7df21
sockets: 2
tablet: 0
vga: qxl
vmgenid: e37435c3-09c6-4069-aefa-bc2328418d91
vmstatestorage: local-lvm

comment: scsi7 is the drive where I try to play around with different zvol blocksizes.



Anyway if IPv4 && v6 are enabled I only get a througput of a 1 Gbit connection (~ 120 Mb/s) independently if I copy stuff over to a NVMe SSD or regular spinner HDD.

As soon as I disable IPv6 within Windows network settings the network throughput gets a lot better (~ 600 Mb/s).


Is this a known bug?