Speed over Samba mount

    • OMV 4.x
    • Speed over Samba mount

      Hello,
      I have installed OMV4 on an odroid.
      I also have an Ubuntu 18.04 as a server with a Raid.

      I am surprised that the speed of the OMV 4 is fast. The Ubuntu system is slower and with bigger files copying from windows, I have to often retry to copy the total amount, because of upcoming windows errors.

      What is different configured in OMV4. I know the extra options. But this do not solve all on the ubuntu system.

      How can I make my ubuntu system working right ?

      Thanks for helping.
      Stefan

      The post was edited 1 time, last by stefluik ().

    • meetthemeat wrote:

      I found this
      forum.openmediavault.org/index…ing-Samba-for-more-speed/
      OMV4 on the ARM devices like ODROIDs uses a bunch of optimizations only partially related to the above smb.conf tuning. And the most important part is this here:

      stefluik wrote:

      with bigger files copying from windows, I have to often retry to copy the total amount, because of upcoming windows errors
      So there's a real problem not related to performance tunables but trouble also impacting performance negatively. This needs to be addresses (by checking logs on the server and performing some basic tests/benchmarks here and there). But without any details it's impossible to help.
    • Hello,
      thanks for response. I have made a Speedtest with LAN-SpeedTest from Totusoft.
      All drives mounted with 6Gb/s SATA.
      Here are the results:

      odroid system is an hc1

      Ubuntu system is an Intel(R) Pentium(R) CPU G630T @ 2.30GHz

      what log-files do you need ?

      Stefan
      Files
    • On the Ubuntu machine: Try without RAID, just a single newly ext4 formatted hard drive and with a fresh Ubuntu install without anything else installed. Less is more! Then try tweaking SMB/CIFS if the performance still is bad. Try tweaking IP packet sizes.
      OMV 4, 7 x ODROID HC2, 1 x ODROID HC1, 5 x 12TB, 1 x 8TB, 1 x 2TB SSHD, 1 x 500GB SSD, GbE, WiFi mesh
    • Source Code

      1. Write Speed: 16,41 MBps
      2. Read Speed: 100,72 MBps
      3. ...
      4. Write Speed: 3,78 MBps
      5. Read Speed: 99,98 MBps
      Both write speeds are way too low. This is not a Samba issue but something more general. I would test with iperf3 (no storage influence) and Helios LanTest (testing a bunch of different stuff). And as usual looking at the problem from bottom to top is a good idea: that's network layer 1 (cabling and Switch) as first measure...
    • I made some tests with helios and iperf3.
      Images
      • odroid helios test.png

        69.78 kB, 614×726, viewed 71 times
      • ubuntu raid helios test.png

        73.56 kB, 614×726, viewed 62 times
      • ubuntu ext4 helios speedtest.png

        68.45 kB, 614×726, viewed 68 times
      Files

      The post was edited 2 times, last by stefluik ().

    • stefluik wrote:

      here are the tests with iperf -R.
      Clearly indicating you having a network problem:

      Source Code

      1. [ 4] 0.00-10.00 sec 200 MBytes 168 Mbits/sec 0 sender
      2. ...
      3. [ 4] 0.00-10.00 sec 253 MBytes 212 Mbits/sec 0 sender
      There's no point in looking at those lousy 212 Mbits/sec the ODROID achieves compared to the even more lousy 168 Mbits/sec of your Ubuntu box. All numbers are way too low and you need to search for the root cause (on network layer 1: cables, connectors, switch).

      There is a real network problem but modern OS partially can mitigate such stuff. This is an interesting read on the strategies file managers of popular client OS implement and why synthetic benchmarks might show lower scores than e.g. Windows Explorer: helios.de/web/EN/support/TI/157.html
    • stefluik wrote:

      what network problem do you think?
      No idea. Usual strategy in such situations is to replace layer 1 and layer 2 hardware until Gigabit Ethernet speeds are reported as exceeding 930 Mbits/sec. I've seem so many strange root causes for bad network performance (including a whole cabinet full of faked network cables or braindead QoS settings on switches trashing performance of everything except babbling -- called VoIP) that it's useless to speculate.

      Exchange cables and switch(es) as first measure...

      The post was edited 1 time, last by tkaiser ().

    • stefluik wrote:

      Hello,
      I am now using a virtualbox Ubuntu on the server to download and to copy the data to the server (same machine) via smb mount.
      after some gb it cancel with an error. why?
      So your VM and your SMB server are running on the same physical machine? If so, you may be thinking that it's bypassing your physical networking equipment entirely, and thus ruling out a network hardware problem. This may not be the case, depending on how you've configured your virtualbox networking. If you're using bridged networking (virtual and physical servers all have IPs on the same subnet) the IP packets are still going out over your network, hitting your router, and coming back to the exact same NIC (but addressed to a different IP). So if you're thinking you've taken network hardware out of the equation - not necessarily.

      I agree that you have a network problem. Replacing cables is the easiest place to start. If you can't easily swap out routers and switches, at least try powering everything down at the same time, and bringing it back up. Routers and even switches can sometimes lose their minds if they run for months and months without a reboot. If you start getting >930 mbps in iperf after restarting the network gear, keep an eye on it - you may find that you need to restart a specific router/switch every month or two (which you might be able to live with). Or you might find you have to do it every day or so (or every hour or so), which might drive you to replacement sooner.