Announcement

Collapse
No announcement yet.

Btrfs Gets Talked Up, Googler Encourages You To Try Btrfs

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

  • nasyt
    replied
    Originally posted by jake_lesser View Post
    I only a few years time, BTRFS will be on every Hard Drive and no one will know what you are talking about if you ask them what ZFS is.
    BTRFS will make Linux even slower as it is.

    Hard-drives used for windows server will have ReFS wich is super fast. It is a shame.

    Leave a comment:


  • 0xBADCODE
    replied
    Originally posted by drSeehas View Post
    I do not think it would happen at all.
    It would not happen this way, but if there will be demand, some day someone can try to create driver like it happened with EXT4. However, btrfs is far more complicated and so it is less likely to happen and would take far longer if someone would dare to pick up such task. Earlier IIRC Apple explicitly rejected idea to support ZFS in MacOS X. I do not know what caused them to make this decision but its really up to them.

    It matters definitely, if you claim, "I only a few years time, BTRFS will be on every Hard Drive".
    It matters for one who makes such claim, but it do not have to matter for others, most notably, devs, FS users and so on.

    And if we're about filesystems for removable storages, MS preferred to be utterly uncooperative. They created exfat which isn't anyhow noteworthy from technical point of view, uses ancient disk techs, etc. Yet it is patent-encumbered and requires to pay royalty. Isn't it cool to pay for ancient and outdated techs? So MS shouldn't complain if others do not want to cooperate with such vendor at all.

    Leave a comment:


  • drSeehas
    replied
    Originally posted by 0xBADCODE View Post
    ... BUT I do not think it would happen this way. ...
    I do not think it would happen at all.

    ... Market share? Well, I do not see why it matters ...
    It matters definitely, if you claim, "I only a few years time, BTRFS will be on every Hard Drive".

    Leave a comment:


  • 0xBADCODE
    replied
    Originally posted by jacob View Post
    MS is developing RTFS whose design and planned features ressemble those of BTRFS. However in its present form (Windows 2012) it is nowhere near being stable or feature complete and I have no idea how it stands in terms of performance.
    I know about it. But...
    1) Its server-only. OTOH I do not see why decent CoW with snapshots is bad for desktops. Just imagine: you've deleted directory with photos by mistake. Or, even worse, overwrote valuable file with another one and it turned out to be NOT what you wanted. Revert to snapshot (or just mount snapshot) - and woo-hoo, data are back! Sure, this implies some overhead for storage of temporary and especially persistent snapshots and differences. However if your disk space allows it, it could be nice idea to use multiple states/views and it comes at cheap price since general idea that only changed data are stored separately while matching blocks are reused (can be further improved via dedup techniques). Should you screw something up, there could be relatively simple "undo" and it does not implies bug performance penalty since that's how CoW works on its own. CoW does not destroys data immediately in most cases, that's what allows to have multiple states.

    2) ReFS features set seems to be really narrow. There are many CoW-like designs around these days. Ok, MS finally got idea their NTFS disk techs from 90s are outdated "a bit", so it's no longer "New Technology" at all . But I do not see any exciting features. And it appeared quickly, so it can happen its internal design haves some silly shortcomings which are hard to undo at later stages, when filesystem is deployed. You see, wide set of features in Btrfs has been planned from begin. And many thinge were taken into account at design phase. So, btrfs can move away data from particular drive to make it unused and remove from pool. This needs back references stored to do it fast and easy. And it would be much better if you create them as you write data, it is hard to do it sometimes later. Then brtfs can enable things like CoW per file and even assumes RAID is block level thing but rather object level, so you can have different RAID levels for different objects. While it is not fully implemented yet (theoretically, you can specify RAID level per file), it still already offers interesting options like storing data and metadata using different RAID levels. And technically underlying design can deal with per-file RAID levels. Needless to say, such things should be taken into account at design phase to be anyhow efficient. That's what makes me suspicious - let's see if MS would be able to extend features at all in sane and efficient ways to be on par. I wouldn't bet on it "by default" granted that MS recently mostly good in marketing and suxx at everything else. Btrfs haves particular name behind design. At least Chris Mason dares to admit he architected this thing. And who is from MS side? Just "MS"? No names of architects, etc? Hmmph, does not looks too promising.

    3) Well, its windows. Its proprietary nature, closed ecosystem, lack of internal details and closed sources are making it exceptionally boring for me. So I havent even bothered myself to learn details about ReFS. I would never use it anyway since I'm not silly enough to step into vendor-locked solutions anymore.

    Originally posted by drSeehas View Post
    As you can read in my post you replied to, I questioned the following statement:
    If Btrfs will be on every hard drive (including SSDs), the standard file system in Windows and MacOS (market share) has to be Btrfs! No 3rd party driver.

    Now you get my point?
    I got it, BUT I do not think it would happen this way. Furthermore, driver model of NT haves nothing to do with Linux, so it's not like if Linux kernel module source too much use under Windows. You'll have to rewrite driver almost completely anyway. Sure, maybe it is possible to copy-paste some code with math/parsing/etc. But again, in Linux people care about solving their tasks in efficient ways. They do not give it a fsck how hard or easy it would be for MS to integrate their code into Windows. That's IMO perfectly fair, since MS does not bothers self too much to share code on sane terms and take care about mutual advantages. Quite similar could be told about Apple as well - their overall attitude hardly can be called opensource-friendly IMO. So it is really wrong for them to complain others could be not in mood to cooperate with them. GPL works well for Linux and it's not like if someone going to change that (and it's technically hard anyway). If you haven't got it, in opensource world, people join their efforts to solve THEIR problems in efficient ways as long as they share ideas about terms and goals. This means, if MS or Apple want to use btrfs - it is very unlikely Linux devs would care about it. So it is up to apple and MS what to do, if they think they need such filesystem as default FS. They can rewrite driver, agree to make driver (or whole kernel, depends if it can be called "linking") GPLed, or do whatever stuff they want. Market share? Well, I do not see why it matters in case of filesystem meant for internal drives and multi-disk storages. That's not a very good choice for filesystem on flash sticks, etc which everyone inserts in each and every computer with arbitrary OS around.

    And filesystem is here and works. Everyone who really needs it can just go use Linux, it costs nothing (unless you're in mood to pay for commercial support), does not puts you to vendor locks, and Linux is not anyhow worse than anything else around, especially in right hands. There is also option someone can write 3rd party driver like it happened in ext2fsd which can mount ext4 in windows. But I wouldn't really count on it. Interestingly, MS aggressively pused filesystem incompatibility at NT times. Its rather funny to see how they could get the very same policy applied to their butts (for different reasons). Gives me a good laugh . "Treat others the way you want to be treated". Something that MS and Apple should learn.

    Leave a comment:


  • jacob
    replied
    Originally posted by Blue Thunder View Post
    RTFS? Do you mean Microsoft's Resilient File System (ReFS): https://en.wikipedia.org/wiki/ReFS
    Yes, of course, ReFS. Sorry.

    Leave a comment:


  • Blue Thunder
    replied
    Originally posted by jacob View Post
    MS is developing RTFS whose design and planned features ressemble those of BTRFS. However in its present form (Windows 2012) it is nowhere near being stable or feature complete and I have no idea how it stands in terms of performance.
    RTFS? Do you mean Microsoft's Resilient File System (ReFS): https://en.wikipedia.org/wiki/ReFS

    Leave a comment:


  • drSeehas
    replied
    Originally posted by 0xBADCODE View Post
    ... So there shouldn't be issues with 3rd party drivers, etc. ...
    As you can read in my post you replied to, I questioned the following statement:
    I only a few years time, BTRFS will be on every Hard Drive ...
    If Btrfs will be on every hard drive (including SSDs), the standard file system in Windows and MacOS (market share) has to be Btrfs! No 3rd party driver.

    Now you get my point?

    Leave a comment:


  • jacob
    replied
    Originally posted by 0xBADCODE View Post
    However, btrfs is generally meant to be multi-disk scalable filesystem and it does not performs very well on small storages below dozen and half gigz. So it likely to be internal filesystem for storages rather than filesystem you use to exchange data between systems. Granted this fact it is just not really important if MacOS or windows can read it. It will be Linux local competitive advantage if they can't. Funny thing is that MS own tactics stabs them in their back :-)
    MS is developing RTFS whose design and planned features ressemble those of BTRFS. However in its present form (Windows 2012) it is nowhere near being stable or feature complete and I have no idea how it stands in terms of performance.

    Leave a comment:


  • 0xBADCODE
    replied
    Originally posted by drSeehas View Post
    The Btrfs license is compatible with Microsoft Windows and Apple MacOS?
    Or do you honestly think, in only a few years time, there will be no longer Windows and MacOS?
    Windows and MacOS X are permitting to have very restrictive EULAs in software, far more restrictive than any GPL and somesuch. So there shouldn't be issues with 3rd party drivers, etc. Say, there is ext2fsd which is GPL-licensed and it mounts EXT4 under Windows. It here and it works. Worse than original driver, but at least you can read/write ext4 drives under windows if you need it.

    However, btrfs is generally meant to be multi-disk scalable filesystem and it does not performs very well on small storages below dozen and half gigz. So it likely to be internal filesystem for storages rather than filesystem you use to exchange data between systems. Granted this fact it is just not really important if MacOS or windows can read it. It will be Linux local competitive advantage if they can't. Funny thing is that MS own tactics stabs them in their back :-)
    Last edited by 0xBADCODE; 08-29-2014, 08:02 PM.

    Leave a comment:


  • Stellarwind
    replied
    Originally posted by jake_lesser View Post
    I only a few years time, BTRFS will be on every Hard Drive and no one will know what you are talking about if you ask them what ZFS is.
    I hear that for years already and yet there is zfs on my hard drives still, which never caused me any issues since very early days of zfs on linux project.

    Leave a comment:

Working...
X