Link Aggregation in Windows 10 Pro

Discussion in 'Network questions and troubleshooting' started by Lavcat, Apr 2, 2022.

  1. Lavcat

    Lavcat Master Guru

    Messages:
    493
    Likes Received:
    32
    GPU:
    AMD Radeon VII
    Thankfully I got my system back together and was able to run some tests. The metric I chose was to compare Macrium Reflect image backup times of my Samsung 980 Pro M.2 drive. The state of the 980 Pro was not quite identical for each backup, however not much data changed between runs.

    1. Backup to locally connected LaCie portable USB disc
    2:46:45

    2. Backup to NAS 1Gb connection before installing VM on the NAS
    4:32:25

    3. Backup to NAS 10Gb connection Windows 11 VM running
    5:07:27

    4. Backup to NAS 10Gb connection VM shut down and Macrium compression off
    4:48:41


    Sadly the bottleneck is not the speed of the connection. Forget link aggregation, a 10Gb connection is no improvement over a 1Gb connection. I feel stupid. I hope Amazon will take the 10Gb cards back. They are too small to use as doorstops.

    As to the Windows 11 VM remote console, the issue was using the default display driver. With an updated driver the remote console is fast and responsive even at high quality and high resolution.

    Amusingly I am still getting errors that I don't have enough bandwidth to support high quality display, even running with the 10Gb NIC's.
     
  2. 386SX

    386SX Ancient Guru

    Messages:
    1,635
    Likes Received:
    1,842
    GPU:
    AMD Vega64 RedDevil
    Why do you test under different conditions with each run? Every time another thing is enabled or disabled. Sorry, but your tests make no sense. Please test the network only, not any software or different test conditions involved, keep it simple. ;)

    Something is holding you back, I am almost 100% sure. Could you please switch off everything not needed during copy (any VMs and stuff) and then copy some big file from your computer to the NAS, like a backup file of some dozen GBs ...?
     
    anticupidon likes this.
  3. anticupidon

    anticupidon Ancient Guru

    Messages:
    6,971
    Likes Received:
    3,264
    GPU:
    Polaris/Vega/Navi
    Tests have to be consistent and run several times, and noted the time.
    Usually I run iperf 5 times when testing a network connection and the value i take is the media between all runs.
     
  4. Lavcat

    Lavcat Master Guru

    Messages:
    493
    Likes Received:
    32
    GPU:
    AMD Radeon VII
    In test 3 I had intended to shut down the VM before running the backup. That was an error on my part. When I thought of it I did not feel like getting up out of bed to rerun the backup.

    However in real life the VM is running 24/7 to provide essential network services. So it doesn't really matter -- except for intellectual interest -- how blazingly fast the backups or file transfers might run with the Windows VM shut down.

    Large file copy is not a typical use case for me. I'd do the experiment you ask except that the network cards are on their way back to Amazon. Hopefully faster than they performed in my system.

    Let me be clear: I have no reason to doubt the 10Gb link speed. The issue is that disappointingly the 10Gb link did nothing to enhance the throughput of my network.

    Possibly I might see more performance by adding more memory to the NAS, but for the time being I am through playing with it. It works. I can live with the speed.

    Just that it's funny I started this thread to be about 10Gb link aggregation when I can't even get throughput above 1Gb from high end 10Gb network cards.
     

  5. Lavcat

    Lavcat Master Guru

    Messages:
    493
    Likes Received:
    32
    GPU:
    AMD Radeon VII
    I have logs of several image backups to my LaCie USB drive. The backup times do not vary much. Why would you expect image backup times to the NAS to vary significantly run to run?
     

Share This Page