Announcement

Collapse
No announcement yet.

Ubuntu's Work On New Desktop Installer Continues, Evaluating ZFS Desktop Support

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Ubuntu's Work On New Desktop Installer Continues, Evaluating ZFS Desktop Support

    Phoronix: Ubuntu's Work On New Desktop Installer Continues, Evaluating ZFS Desktop Support

    A few things in Ubuntu's latest weekly development summary caught our attention... As has been going on for months, a new Ubuntu installer "Ubiquity-NG" continues to be worked on, but seemingly tying into that they are looking at ZFS support on the desktop...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Don't know what's the point supporting zfs or any other exotic filesystems while they can't create bcache during installation.

    Comment


    • #3
      Originally posted by Khrundel View Post
      Don't know what's the point supporting zfs or any other exotic filesystems while they can't create bcache during installation.
      Meh, i'm hoping bcachefs lives up to the promise (of which btrfs failed todo) then we won't have to worry about ZFS.

      Comment


      • #4
        Meh, another abandoned canonical project just started. :->

        Comment


        • #5
          Originally posted by Khrundel View Post
          Don't know what's the point supporting zfs or any other exotic filesystems while they can't create bcache during installation.
          Because ZFS has a lot of advanced features that need to be fleshed out in order to be Linux optimized.

          One example is databases. By default, ZFS kind of sucks with databases. Tweak the recordsize and how the cache is handled and it has a noticeable performance boost on databases. My method is to deal with that is to setup a special database datasets (one for every DB format) and either symlink or mount databases to that dataset.

          Another one is torrents. They can fragment the hell out of ZFS as well as need recordsize tweaks to be ZFS optimized. My solution to that is a special /torrent directory/dataset with the correct recordsize and then moving the downloaded file form /torrent to /where/I/need/it on another dataset.

          While that is just two examples, it's only to show that advanced filesystems aren't like EXT4 -- you don't format and go with it. Depending on the type of data one is going to manipulate or what they're doing, the default settings suck. ZFS has more advanced features than any other Linux filesystem and a lot of the tweaks needed for ZFS actually do translate over to BTRFS and bcachefs.

          Treating ZFS as a first class citizen now means all the bits and pieces will be in place for a well optimized system when other filesystems finally catch up to ZFS.

          Comment


          • #6
            Hmm, interesting. No "but why another installer when they can use a more standardized one like Anaconda" posts???

            Comment


            • #7
              Originally posted by Vistaus View Post
              Hmm, interesting. No "but why another installer when they can use a more standardized one like Anaconda" posts???
              I guess everyone knows this is what Canonical does by now. No sense in asking over and over.

              Comment


              • #8
                Does this mean they are going to pull ZFS into their own kernels?

                One would hope, otherwise having your root storage reliant on DKMS is just asking for trouble.

                Comment


                • #9
                  Doubtful, Oracle's attorney arm will rape them.
                  About ZoL,ZoF..its not "zfs" until the OS lacks "
                  beadm" or similar command.

                  Comment


                  • #10
                    Originally posted by edenist View Post
                    Does this mean they are going to pull ZFS into their own kernels?

                    One would hope, otherwise having your root storage reliant on DKMS is just asking for trouble.
                    No. Only end users can legally do that. The whole CDDL/GPL thing...

                    It really isn't as bad as it sounds. I (used to) do it. Once ZFS hit 0.8.0rc and the module merge happened, it quit being a worry as long as you use a kernel the ZoL project says is compatible, 2.6.32 to 4.20.x with ZoL stable (2.6.32 to 5.0rc with ZoL git master). I've never had ZFS fail to build on a supported kernel outside DKMS not building SPL and ZFS in the correct order in over 4 years now -- it's pretty reliable is what I'm saying.

                    ZFS and DKMS really isn't an issue for distributions like Ubuntu and Manjaro. They'll ship the module in a package and the end user won't have to compile it themselves. I still use the DKMS version, but only because I have 0.8.0rc formatted zpools. I'll switch back to the distro provided module whenever 0.8 is released.

                    I mention Manjaro because that is how Manjaro does it now. Been a while since I've used Ubuntu, but I have to imagine they'd easy-mode ZFS like Manjaro does.
                    Last edited by skeevy420; 06 February 2019, 08:58 PM.

                    Comment

                    Working...
                    X