Announcement

Collapse
No announcement yet.

Linux 5.18 Looks Like It Will Finally Land Btrfs Encoded I/O

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

  • Linux 5.18 Looks Like It Will Finally Land Btrfs Encoded I/O

    Phoronix: Linux 5.18 Looks Like It Will Finally Land Btrfs Encoded I/O

    Btrfs file-system developers remain very busy on multiple fronts from performance and reliability enhancements to in-development Btrfs on-disk format changes. A notable item now on deck for integration in Linux 5.18 is Btrfs support for encoded I/O...

    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
    As a user, I am very pleased with how actively Btrfs is being developed. But I still hope that problems with RAID 5/6 will be addressed.

    Comment


    • #3
      Ok, so if a user has full disk encryption and a program decides to use this API, it can write unencrypted data?
      With many governments trying to ban or backdoor encryption, it seems to me that with such a system in place they just need to force the programs to write unencrypted data to disk even though the user is thinking that everything will be encrypted.
      Please tell me that this is not the case as I find it the timing as a strange coincidence!

      Comment


      • #4
        It's about data compression.

        Comment


        • #5
          Originally posted by Eraserstp View Post
          As a user, I am very pleased with how actively Btrfs is being developed. But I still hope that problems with RAID 5/6 will be addressed.
          Honestly it's unlikely to happen anytime soon. Those who invest in btrfs development (FB, SUSE...) don't care about RAID 5/6. On the other hand, those who do care (small companies, home NAS users etc) aren't prepared to actually do the work or fund it. Catch 22?

          Comment


          • #6
            Originally posted by jacob View Post

            Honestly it's unlikely to happen anytime soon.
            I know this, but I want to hope that the problem will be solved sometime, the more popular the file system, the more users using it and the more developers working on it.

            Comment


            • #7
              Originally posted by Eraserstp View Post

              I know this, but I want to hope that the problem will be solved sometime, the more popular the file system, the more users using it and the more developers working on it.
              In practice someone might start working on it if/when btrfs gets adopted by default by a player in the small/home server or NAS space. Facebook has no motivation to do it, SUSE is enterprise oriented; Fedora uses it in its desktop distros but that's not fertile ground for RAID 5/6, even less the IoT or embedded flavours. Don't count on Canonical either, they are all about the cloud. Ideally the necessary effort might come from the likes of Debian or Arch, but those are community distros with few resources and they can't and won't take on upstream development. That leaves us with someone like Synology, but they have hardly ever contributed anything upstream and their use case is already served adequately anyway.

              Comment


              • #8
                Originally posted by jacob View Post

                Honestly it's unlikely to happen anytime soon. Those who invest in btrfs development (FB, SUSE...) don't care about RAID 5/6. On the other hand, those who do care (small companies, home NAS users etc) aren't prepared to actually do the work or fund it. Catch 22?
                I find this quite ironic because ZFS which also targets enterprise did address RAID 5/6 in its initial design even though back then no one was really running ZFS on home NAS (in fact back then the difference was even more extreme then it is now).

                Also its good to qualify why enterprise doesn't use RAID 5/6. Its not because it doesn't fulfill a legitimate usecase (there are definitely usecases where you want to optimize more for storage then for raw throughput which is why RAID 5/6 exists) but rather most implementations of RAID 5/6 were horrible/buggy because its harder to implement correctly, especially so if you use hardware RAID which was very typical up until somewhat recently. Due to this, historically most serious data centers just avoided RAID 5/6 like the plague and instead just opted for RAID 10 and accepted the extra hard drives as a necessary cost.

                Arguably ZFS's implementation of RAID 5/6 (which they call raidz 1/2/3) was one of the first comprehensive/correct/working implementations and if ZFS on Linux really picks up I wouldn't be surprised if a lot more people use raidz 1/2/3.
                Last edited by mdedetrich; 11 February 2022, 02:43 AM.

                Comment


                • #9
                  Originally posted by mdedetrich View Post
                  I find this quite ironic because ZFS which also targets enterprise did address RAID 5/6 in its initial design even though back then no one was really running ZFS on home NAS (in fact back then the difference was even more extreme then it is now).
                  That was back when RAID-5/6 was more common in enterprises. I think most cloud and hyperscalers are probably now using distributed filesystems with replication, and lots of enterprises have migrated their back office to the cloud.

                  One reason traditional RAID is losing favor is due to rebuild times. But it also just doesn't scale as well.

                  Comment


                  • #10
                    This could improve package installation in Fedora even further as decompression+recompression could be skipped completely if I see it correctly:

                    Comment

                    Working...
                    X