Announcement

Collapse
No announcement yet.

Fedora 16 May Default To Btrfs File-System

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

  • #21
    Originally posted by devsk View Post
    Have a look at zfs on linux page I linked to in my previous post. It is portable enough, because its working already! I am using zfs on linux on my desktop/server.
    Did you do some benchmarks how it compares to Ext4 on Linux?

    Comment


    • #22
      Originally posted by curaga View Post
      @devsk

      I'm curious, is ZFS like btrfs in the aspect that it uses obscene amounts of space for its metadata?

      That is pretty much why I won't be using btrfs, among the other reasons.
      If you saw my bug report, I had the exact same concern. Here is what I did. I have an external disk as a backup of backup. So, I went ahead and created BTRFS (we are talking 2.6.38.8) on it and rsynced my ZFS backup onto it. The resulting disk space usage was much better on the ZFS compared to BTRFS. I can provide the real numbers when I go home later. But ZFS was the clear winner in terms of space efficiency. Both filesystems used their default compression (which gives advantage to BTRFS because it uses much stronger zlib compression by default, but it still ended up eating more space). This is a good test case because it had variety of data because its a backup from multiple machines.

      I have not been able to kill a ZFS filesystem so far, no matter how hard I try, not on OpenSolaris, not on zfs-fuse and not on zfs-on-linux. And best part is that it has detected 2 bad disks in last 1.5 yrs of my usage, without killing my FS. Compare that to BTRFS: my rootfs (back when I was a BTRFS zealot, I ran rootfs on BTRFS, now the idea seems crazy!) had corrupted itself (no power loss, no reset) but BTRFS was happy about it. It would list those files with garbage characters in filenames, I couldn't read them, btrfsck couldn't rectify them and kernel was happy about it (read [1]). It was either a disk issue or just a genuine FS corruption. But what good are checksums if the FS can't detect faults using them? That's why I say BTRFS just claims checksum support but just having checksums doesn't mean anything. Putting them to some good use is THE feature!

      [1] http://forums.gentoo.org/viewtopic-p...s.html#6280748

      Comment


      • #23
        Originally posted by kraftman View Post
        Did you do some benchmarks how it compares to Ext4 on Linux?
        I wish raw filesystem speed was everything in life! Its not! I won't fall for it ever! A faster FS eating my data is no good for me.

        But yes, I did do comparisons. And ZFS on Linux is beaten in almost all of them except a few. Random access is much better, space is more efficiently used, read scaling with more drives is good.

        Note that ZFS on Linux is a baby at this time. The layers which were added as glue between ZFS code and Linux kernel need to become mature. It will improve in performance with time. But it has a solid foundation to base that on. Needless to say, I am very excited about this project!

        Comment


        • #24
          yyyyyyyyyyeah a filesystem is one of those things where stability and reliability are kinda important. I mean if my mplayer crashes every now and then (it doesn't, just an example) I might throw a hissy fit or two, but in the end it's no big deal. Filesystem stability is one of the biggest requirements for an OS distribution and definitely a "non-negotiable" for MOST uses, but especially in those uses where these advanced features actually matter.

          With ext and ntfs you have many, many years of testing and development under the belt. ZFS has been around awhile too. I might have to steer clear of btrfs for at least the next five years... lol.

          Comment


          • #25
            Originally posted by devsk View Post
            I wish raw filesystem speed was everything in life! Its not! I won't fall for it ever! A faster FS eating my data is no good for me.

            But yes, I did do comparisons. And ZFS on Linux is beaten in almost all of them except a few. Random access is much better, space is more efficiently used, read scaling with more drives is good.
            I'm just asking, because I'm quite sceptic in using btrfs (and ZFS) on a simple desktop. I probably won't use 90% of its capabilities, so while Ext4 is faster at this point I'll stick with it for now.

            It will improve in performance with time. But it has a solid foundation to base that on. Needless to say, I am very excited about this project!
            That for sure!

            Comment


            • #26
              Originally posted by kraftman View Post
              I'm just asking, because I'm quite sceptic in using btrfs (and ZFS) on a simple desktop. I probably won't use 90% of its capabilities, so while Ext4 is faster at this point I'll stick with it for now.
              Makes sense. My rootfs is ext4 as well and it has served me well. Its not changing anytime soon.

              Comment

              Working...
              X