Announcement

Collapse
No announcement yet.

Btrfs File-System Changes Published For Linux 3.13

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

  • phoronix
    started a topic Btrfs File-System Changes Published For Linux 3.13

    Btrfs File-System Changes Published For Linux 3.13

    Phoronix: Btrfs File-System Changes Published For Linux 3.13

    The Linux 3.13 kernel brings major enhancements to Samsung's F2FS file-system but the EXT4 and XFS changes aren't too exciting. How are the Btrfs changes for this next kernel? We now know thanks to a new pull request from Chris Mason...

    http://www.phoronix.com/vr.php?view=MTUxNDg

  • ulenrich
    replied
    large random access file - like vm swapfile etc

    I never understood how this works with cow and snapshot functionality?

    Are there exceptions declared by filename?
    Is there a COW_MAX_FILESIZE defined in /etc or in the btrfs source?

    Leave a comment:


  • jwilliams
    replied
    Originally posted by smitty3268 View Post
    Yep, and people/distros are finally starting to use it. Once that happens...
    No major distros are installing btrfs as the default filesystem. A few keep saying they will soon, but soon never seems to happen. btrfs just is not ready. It seems directionless. It lacks good project management.

    Leave a comment:


  • smitty3268
    replied
    Originally posted by jwilliams View Post
    People have been saying that for several years now
    Yep, and people/distros are finally starting to use it. Once that happens, I think the remaining problems will probably clear up pretty quickly because people will start caring about them a lot more than they do right now.

    Anyway, i don't get the big deal. Use whatever FS you want. Who cares? Btrfs will either get better and people will start using it. Or it won't. End of story.

    Leave a comment:


  • jwilliams
    replied
    Originally posted by smitty3268 View Post
    It's not quite ready yet, but let's wait another year and see how things stack up then.
    People have been saying that for several years now

    Leave a comment:


  • BSDude
    replied
    Originally posted by jwilliams View Post
    I would not count on btrfs being "finalized" any time soon.

    The btrfs project was never very focused and directed. Now that Mason no longer works for Oracle, the project seems even more directionless than before. There are a number of people fixing bugs, but there always seem to be more bugs being discovered, and the important issues do not seem to get fixed or completed (qgroups, free space, parity RAID, snapshots, etc.)
    I think it also has to do with the fact that Oracle bought Sun and "owns" ZFS. Maybe they're focused on it more than on BTRFS? From one perspective it's good to have alternatives but then why spend resources when you have an advanced filesystem that does everything BTRFS was designed to do, especially when you're Oracle and are trying to squeeze every penny out.

    Leave a comment:


  • smitty3268
    replied
    Originally posted by jwilliams View Post
    Unfortunately, btrfs is worse than ext4 in several ways, primarily reliability and free space issues, but also it is a poor choice for a swap file or for a large, random access file like that for a VM.
    Sure. It's also better in several ways.

    It's not quite ready yet, but let's wait another year and see how things stack up then.

    Leave a comment:


  • jwilliams
    replied
    Originally posted by elmystico View Post
    Luckily with btrfs I can convert my raid10 volumes to raid5 when time comes, on-the-fly :-)
    If you are "lucky", the time might come in another 6 years or so.

    Leave a comment:


  • elmystico
    replied
    From my point of view biggest missing features are lz4 compression and proper raid.

    lz4 compression- I'm using some few-years-old systems and lzo effectively chokes the hardware but gives some additional space of course. Now, few weeks ago I've started using zswap with lz4 compression- and it's a big win even with old intel atom 1,6 GHz processor, with few percents of processor utilization. I guess that having it with btrfs would give huge boost for transfer speed of stored data with decent space savings. (zswap uses some compressed RAM space for swap and block devices are slower than ram, right?) It's like a huge technological breakthrough, I guess this kind of compression can be turned on on 95% of systems safely, by default, giving bigger storage space AND performance at the same time.

    raid thoughts- performance is quite inferior to md-raid, some of the reasons are from missing functionalities in btrfs raid implementation. For example no different data ordering options for raid10 (far, offset etc), no selection of chunk size (? I'm not sure), no odd drives number. As of raid5- I'm to scared to even try if this works for now, as far as I know it's rather incomplete. Luckily with btrfs I can convert my raid10 volumes to raid5 when time comes, on-the-fly :-)

    Leave a comment:


  • GreatEmerald
    replied
    Originally posted by jwilliams View Post
    False. You cannot "restore snapshots" without rebooting when certain changes have been made. For example, different kernel.
    Guess what, when you update your kernel, you have to reboot for it to take effect. There is no way around that (aside from kexec, but that's also not very reliable). And it's not the filesystem's job to deal with problems like that.

    Leave a comment:

Working...
X