XCP-ng
    • Categories
    • Recent
    • Tags
    • Popular
    • Users
    • Groups
    • Register
    • Login

    How it works the NBD? I don't see improvements

    Scheduled Pinned Locked Moved Xen Orchestra
    7 Posts 2 Posters 393 Views 2 Watching
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • M Offline
      marcom
      last edited by

      Hi all,
      sorry I'm not expert and I would lik to ask you clarifications regarding the network protocol NBD.
      I'm using the last version of Orchestra (update today) from source, and I'm testing it with Xen Server 8.2.

      I did a full backup of one VM and it took 27 Min.
      • Duration: 27 minutes
      • Size: 99.87 GiB
      • Speed: 63.32 MiB/s

      Than I did it again with NBD (4 connection) and the result is the same.
      • Duration: 27 minutes
      • Size: 99.87 GiB
      • Speed: 63.25 MiB/s

      I can confirm the NBD because now I see also the text (Transfer data using NBD).

      CBT I can't test for the moment because the XenServer has only the standard license.

      Can someone explain me a bit more in detail how it works?

      Many thanks in advance

      BR
      Marco

      1 Reply Last reply Reply Quote 0
      • olivierlambertO Offline
        olivierlambert Vates 🪐 Co-Founder CEO
        last edited by olivierlambert

        It's sounds very very similar (maybe too similar), double check the history of CPU usage on your dom0/host with and without it. VHD export is like 20/25% CPU vs 5% with NBD. It could be also another bottleneck somewhere else, as it's a stream, it will max the speed of the slowest element in the chain.

        M 1 Reply Last reply Reply Quote 0
        • M Offline
          marcom @olivierlambert
          last edited by

          @olivierlambert Thanks a lot for the quick reply. I checked with the command "xentop" on my server.

          When I run the Backup, with or without the NDB the CPU increase from 50% to 120-160%.
          9c88f50f-d7e4-44ec-b7a7-acb41411d069-image.png

          Now I'm testing one backup per time, and what I see is that in both scenarios the transfer rate is: 110.18 MiB/s so almost the 1 GB. SO I think the bottleneck is the network. I mean on the Hypervisor I have 2 GB because I have the LACP, but the Virtual NIC of the XO is 1G.
          b15bf703-5465-4c82-8c2b-95d2c3419c67-image.png

          Did I unterstand correctly?

          Thanks

          BR
          Marco

          1 Reply Last reply Reply Quote 0
          • olivierlambertO Offline
            olivierlambert Vates 🪐 Co-Founder CEO
            last edited by

            You can just compare the stats in XO in the host view/stats and CPU graph.

            M 1 Reply Last reply Reply Quote 0
            • M Offline
              marcom @olivierlambert
              last edited by

              @olivierlambert below you see two bacups on the stat, The first is with NDB and the second without.

              6aec3b66-84c3-49a9-b9a8-7fd849cbbacd-image.png

              1 Reply Last reply Reply Quote 0
              • olivierlambertO Offline
                olivierlambert Vates 🪐 Co-Founder CEO
                last edited by

                It's hard to see anything due to the high number of CPUs. It sounds really really similar though, I'm under the impression both are using NBD 🤔 @florent any idea?

                M 1 Reply Last reply Reply Quote 0
                • M Offline
                  marcom @olivierlambert
                  last edited by

                  @olivierlambert If you want me to check anything else, just tell me and I will test for you.

                  Many Thanks

                  BR
                  Marco

                  1 Reply Last reply Reply Quote 0
                  • First post
                    Last post