Announcement

Collapse
No announcement yet.

Western Digital's Zonefs File-System Looks Like It Could Be Ready To Land With Linux 5.6

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

  • Western Digital's Zonefs File-System Looks Like It Could Be Ready To Land With Linux 5.6

    Phoronix: Western Digital's Zonefs File-System Looks Like It Could Be Ready To Land With Linux 5.6

    Introduced last month was Zonefs as a new Linux file-system developed by Western Digital. It's looking like that new file-system could be ready for introduction with the upcoming Linux 5.6 cycle...

    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
    Any organization willing to deploy and use it? It would be interesting to know it.

    CERN? NASA? Google?

    Despite this seems only for Big Iron (high end servers, datacenters), it's interesting to know about it.

    Comment


    • #3
      What's the license?

      Comment


      • #4
        Originally posted by kaprikawn View Post
        What's the license?
        Looks like GPL 2.0



        Code:
        diff --git a/fs/zonefs/super.c b/fs/zonefs/super.c
        new file mode 100644
        index 000000000000..94d3f2883d39
        --- /dev/null
        +++ b/[URL="https://git.kernel.org/pub/scm/linux/kernel/git/dlemoal/zonefs.git/tree/fs/zonefs/super.c?h=for-5.6&id=5bba4a0d475aece7213224652fe93f05130aba3a"]fs/zonefs/super.c[/URL]
        @@ -0,0 +1,1177 @@
        
        +// SPDX-License-Identifier: GPL-2.0
        +/*
        + * Simple zone file system for zoned block devices.
        + *
        + * Copyright (C) 2019 Western Digital Corporation or its affiliates.
        + */

        Comment


        • #5
          Originally posted by kaprikawn View Post
          What's the license?
          Is it necessary to ask this for a mainlined module?

          Or are you hoping BSDs can get this filesystem?

          Comment


          • #6
            SMR is terrifying.

            You basically have disk firmware preforming OS invisible non-atomic overwrites to the disk.. not a fan of that, at all. This isn't the direction I'd like to see the industry go and would recommend against their use in enterprise.

            They are the Boeing 737-MAX of storage.
            Last edited by k1e0x; 20 January 2020, 06:20 PM.

            Comment


            • #7
              Originally posted by k1e0x View Post
              SMR is terrifying.

              You basically have disk firmware preforming OS invisible non-atomic overwrites to the disk.. not a fan of that, at all. This isn't the direction I'd like to see the industry go and would recommend against their use in enterprise.
              Aaaand that's another guy talking of the consumer-grade Drive-managed SMR drives that "disguise themselves as normal drives" by firmware.

              FYI: that's how any block device flash media (Flash USB, SDcards and SSDs) and "hardware" RAID cards used in enterprise work.

              And that's definitely the way the industry is going (with NVMe) even if SMR is a niche technology that will probably die off in a few years when flash media takes over.

              There is some push towards the so-called "open-channel SSD" where the firmware isn't disguising itself as a hard drive anymore and relies on the OS to deal with the actual hardware, which is similar to businness-grade Host-managed SMR drives, the main targets of filesystems like the one developed in the article here.

              Comment


              • #8
                Originally posted by starshipeleven View Post
                Aaaand that's another guy talking of the consumer-grade Drive-managed SMR drives that "disguise themselves as normal drives" by firmware.

                FYI: that's how any block device flash media (Flash USB, SDcards and SSDs) and "hardware" RAID cards used in enterprise work.

                And that's definitely the way the industry is going (with NVMe) even if SMR is a niche technology that will probably die off in a few years when flash media takes over.

                There is some push towards the so-called "open-channel SSD" where the firmware isn't disguising itself as a hard drive anymore and relies on the OS to deal with the actual hardware, which is similar to businness-grade Host-managed SMR drives, the main targets of filesystems like the one developed in the article here.
                And that's another guy saying "We don't need to re-certify the plane.. we can make fly exactly the same as the old one using firmware!"

                Willing to bet your life on that code? No? just your data tho?

                SMR is particular hairy as every write includes a delete and rewrite.

                Comment


                • #9
                  Originally posted by k1e0x View Post
                  Willing to bet your life on that code? No? just your data tho?
                  Can you please stop posting this bullshit? memory storage devices never ever had anything remotely resembling the safety level required to trust your life to.

                  In other words, if I had trusted my own life into conventional hard drives I would have died around 6 times by now, and I would be on the verge of dying again in a few months if I trust the SMART reports of my NAS. I don't see how SMR or even flash media change that.

                  Backups, motherfucker.

                  SMR is particular hairy as every write includes a delete and rewrite.
                  It's not more hairy than any flash-based storage, where every write includes a delete and rewrite of a whole 128kbyte or more (that's the erase block size).

                  Comment

                  Working...
                  X