[SOLVED] Migration to a 10 Gbps network

[Blacktime2]

Hi everyone,

I am in the process of migrating my network to 10 Gbps, however, I am experiencing data rate problems.
My host works fine at 10G :



but my VMs stay at 1 Gbps :



I have only vitio network adaptater.
I tried doing an iperf between my VM and the host and this is what I get :



I also tried on another Windows server VM but this time I am limited to 1 Gbps via iperf.
Do you have any suggestions?
Thank you very much for your help!

 

[Richard]

Hi everyone,

I am in the process of migrating my network to 10 Gbps, however, I am experiencing data rate problems.
My host works fine at 10G :

View attachment 44350

but my VMs stay at 1 Gbps :

View attachment 44351

I have only vitio network adaptater.
I tried doing an iperf between my VM and the host and this is what I get :

View attachment 44352

I also tried on another Windows server VM but this time I am limited to 1 Gbps via iperf.
Do you have any suggestions?
Thank you very much for your help!

View attachment 44353

When running the speedtest from ookla you have much more complexity compared with running just iperf3 and it's more difficult to detect the bottlemneck. AFAIU in the current case there is mainly the question if VM versus host is ok (respectively if VM is able to use the host's 10GBps capability). Therefore I would connect locally a server to your 10GBps LAN and then try iperf3 between VM and this server and compare the results with that between host and this server. Using bridged connection compared with routed connection may cause different results. A possible bottleneck is the system performance, especially CPU, can be checked by having a look at node's "Summary" in Proxmox WEB GUI.

 

[Blacktime2]

When running the speedtest from ookla you have much more complexity compared with running just iperf3 and it's more difficult to detect the bottlemneck. AFAIU in the current case there is mainly the question if VM versus host is ok (respectively if VM is able to use the host's 10GBps capability). Therefore I would connect locally a server to your 10GBps LAN and then try iperf3 between VM and this server and compare the results with that between host and this server. Using bridged connection compared with routed connection may cause different results. A possible bottleneck is the system performance, especially CPU, can be checked by having a look at node's "Summary" in Proxmox WEB GUI.

Thanks for your answer.
I cannot run iperf between my VM and another host because the rest of my network is not 10 Gbps. However, I observed my server monitor during the speedtest (LaFibre.info (10 Gbps server)) and the CPU does not exceed 20%.

Edit : I just noticed that when I do a speedtest via the web browser using the same server (Ookla - LaFibre.info) the result is different. In the web browser the speedtest result is 600 Mbps and in the CLI 4 Gbps which is much better. It's the same for Windows or Linux VM.