Hi all,

 

I'm facing some performances issues between Unraid and VMs.

I've the following setup:

• HW : HP MicroServer Gen8
• CPU : Intel® Xeon® CPU E3-1265L V2 @ 2.50GHz
• Memory: 16 GiB DDR3 Single-bit ECC
• OS: Unraid 6.12.6 (Linux 6.1.64-Unraid x86_64)
• eth0: ASUS XG-C100C (AQC113CS NBase-T/IEEE 802.3bz Ethernet Controller [AQtion]) - 10Gbit/s
• etho in br0 (no bonding)
• VLANs: 1, 10

 

VM

• Machine: Q35-7.2
• 8 logical CPUs assigned
• Memory: 4 GB
• OS: Pfsense [2.7.2-RELEASE]
• eth0: Virtio: br0.10 - WAN
• eth1: Virtio: br0 - LAN

 

For the test scenario I 've one VM only and turned off Docker engine.

 

Some iperf3 results

• PC to Unraid: 6.55 Gbits/sec
• Unraid to Unraid: 28.0 Gbits/sec
• Unraid to VM: 831 Mbits/sec
• PC to VM: 1.33 Gbits/sec

 

While I can explain the 6 Gbits/s between PC and Unraid (because cable quality), I cannot explain the results for the internal networking between Unraid (host) and the running VM.

I've read that ideally for pfsense it's better a passthrough NIC, but I cannot add more 10 Gbps interfaces on this HW so I'm forced to share between VMs and Docker with br0.

 

I already tried the switch between virtio to virtio-net, tried to disable TSO and GSO, but without improvements.

 

Any suggestion is really appreciated.

 

Thanks,

Andrea

same issue here, windows sucks bigly so might just be driver issues. Id test with a unix system and see if you see improvements, im like 99% sure the slowness in VM is all windows being stupid. 

Edit: Yes I set up iPerf on Unraid host, test from Windows 11 VM, 1.5gbps.  I set up another iPerf instance on the same network the VM is on and tested between them, Im bonding so I get 40gbps 1 way and 35gbps the other so its not a network issue, its a VM issue.

Edit2: I fixed it by isolating CPU cores for the VM. Getting full 40gbps to any host via iPerf. Interesting that isolating cores was the solution.

Edited October 5, 20241 yr by Jclendineng
Added iPerf details