Hello, Somewhat off topic, but it seems that someone released a COW file system for Linux (currently in ''alpha''): * Extent based file storage (2^64 max file size) * Space efficient packing of small files * Space efficient indexed directories * Dynamic inode allocation * Writable snapshots * Subvolumes (separate internal filesystem roots) - Object level mirroring and striping * Checksums on data and metadata (multiple algorithms available) - Strong integration with device mapper for multiple device support - Online filesystem check * Very fast offline filesystem check - Efficient incremental backup and FS mirroring http://lkml.org/lkml/2007/6/12/242 http://oss.oracle.com/~mason/btrfs/ Via Storage Mojo: http://storagemojo.com/?p=478 -- David Magda <dmagda at ee.ryerson.ca> Because the innovator has for enemies all those who have done well under the old conditions, and lukewarm defenders in those who may do well under the new. -- Niccolo Machiavelli, _The Prince_, Chapter VI
it''s about time. this hopefully won''t spark another license debate, etc... ZFS may never get into linux officially, but there''s no reason a lot of the same features and ideologies can''t make it into a linux-approved-with-no-arguments filesystem... as a more SOHO user i like ZFS mainly for it''s COW and integrity, and being able to add more storage later on. the latter is nothing new though. but telling the world "who needs hardware raid? software can do it much better" is a concept that excites me; it would be great for linux to have something like that as well that could be merged into the kernel without any debate. On 6/14/07, David Magda <dmagda at ee.ryerson.ca> wrote:> Hello, > > Somewhat off topic, but it seems that someone released a COW file > system for Linux (currently in ''alpha''): > > * Extent based file storage (2^64 max file size) > * Space efficient packing of small files > * Space efficient indexed directories > * Dynamic inode allocation > * Writable snapshots > * Subvolumes (separate internal filesystem roots) > - Object level mirroring and striping > * Checksums on data and metadata (multiple algorithms available) > - Strong integration with device mapper for multiple device support > - Online filesystem check > * Very fast offline filesystem check > - Efficient incremental backup and FS mirroring
On June 14, 2007 3:57:55 PM -0700 mike <mike503 at gmail.com> wrote:> as a more SOHO user i like ZFS mainly for it''s COW and integrity, andhuh. As a SOHO user, why do you care about COW? -frank
because i don''t want bitrot to destroy the thousands of pictures and memories i keep? i keep important personal documents, etc. filesystem corruption is not a feature to me. perhaps i spoke incorrectly but i consider COW to be one of the reasons a filesystem can keep itself in check, the disk write will be transactional and guaranteed if it says successful, correct? i still plan on using offsite storage services to maintain the physical level of redundancy (house burns down, equipment is stolen, HDs will always die at some point, etc.) but as a user who has had corruption happen many times (FAT32, NTFS, XFS, JFS) it is encouraging to see more options that put emphasis on integrity... On 6/14/07, Frank Cusack <fcusack at fcusack.com> wrote:> On June 14, 2007 3:57:55 PM -0700 mike <mike503 at gmail.com> wrote: > > as a more SOHO user i like ZFS mainly for it''s COW and integrity, and > > huh. As a SOHO user, why do you care about COW? > > -frank >
On June 14, 2007 4:40:18 PM -0700 mike <mike503 at gmail.com> wrote:> because i don''t want bitrot to destroy the thousands of pictures and > memories i keep?COW doesn''t stop that.> i keep important personal documents, etc. filesystem > corruption is not a feature to me. perhaps i spoke incorrectly but i > consider COW to be one of the reasons a filesystem can keep itself in > check, the disk write will be transactional and guaranteed if it says > successful, correct?Yes, but there are many ways to get transactions, e.g. journalling. -frank ps. top posting especially sucks when you ask multiple questions.
On 6/14/07, Frank Cusack <fcusack at fcusack.com> wrote:> Yes, but there are many ways to get transactions, e.g. journalling.ext3 is journaled. it doesn''t seem to always be able to recover data. it also takes forever to fsck. i thought COW might alleviate some of the fsck needs... it just seems like a more efficient (or guaranteed?) method of disk commitment. but i am speaking purely from the sidelines. i don''t know all the internals of filesystems, just the ones that have bitten me in the past.> ps. top posting especially sucks when you ask multiple questions.yes, sir!
On June 14, 2007 5:07:39 PM -0700 mike <mike503 at gmail.com> wrote:> On 6/14/07, Frank Cusack <fcusack at fcusack.com> wrote: >> Yes, but there are many ways to get transactions, e.g. journalling. > > ext3 is journaled. it doesn''t seem to always be able to recover data.zfs is COW. it isn''t always able to recover data. Like ext3, this is the result of bugs, not a COW vs journalling thing. Or perhaps in a few ext3 cases, the result of bad data making it to disk ... but still not related to COW vs journalling.> it also takes forever to fsck.the journal is supposed to pretty much eliminate fsck time. you might be doing a full fsck as opposed to journal playback.> i thought COW might alleviate some of the fsck needs...It does, but that''s not unique to COW.> it just seems like a more efficient (or guaranteed?) > method of disk commitment.it might be more efficient. i *think* this depends on your usage pattern. i think that "typically", it does outperform other methods. anyway, my point is that i didn''t think COW was in and of itself a feature a home or SOHO user would really care about. it''s more an implementation detail of zfs than a feature. i''m sure this is arguable.>> ps. top posting especially sucks when you ask multiple questions. > > yes, sir!heh -frank
On Thu, Jun 14, 2007 at 17:19:18 -0700, Frank Cusack wrote: : anyway, my point is that i didn''t think COW was in and of itself a feature : a home or SOHO user would really care about. it''s more an implementation : detail of zfs than a feature. i''m sure this is arguable. I''m really not sure I agree with you given the way you''ve put it. Yes, the average user doesn''t give a damn whether his filesystem is copy-on-write, journalled, or none of the above if you ask him. Your average user *does* care, however, when faced with the consequences, and *will* complain when things break. CoW does appear to have some distinct advantages over other methods when faced with the unreliable hardware we all have to deal with these days. When software makes it better, everybody wins. Now all I need is a T2000 that behaves correctly rather than a T2000 that behaves as if it''s missing the magic in /etc/system that it has... -- Dickon Hood Due to digital rights management, my .sig is temporarily unavailable. Normal service will be resumed as soon as possible. We apologise for the inconvenience in the meantime. No virus was found in this outgoing message as I didn''t bother looking.
Darren.Reed at Sun.COM
2007-Jun-22 22:27 UTC
[zfs-discuss] Btrfs, COW for Linux [somewhat OT]
mike wrote:> it''s about time. this hopefully won''t spark another license debate, > etc... ZFS may never get into linux officially, but there''s no reason > a lot of the same features and ideologies can''t make it into a > linux-approved-with-no-arguments filesystem...Well, there''s a dark horse here called "patents". Nobody really knows the full extent of who''s got what or what covers what between the likes of Sun (ZFS/Stotek), NetApp (WAFL) and IBM (ARC?). Maybe this patent game will just end up being like the cold war with nuclear missiles...and the only real winners being the lawyers and patents office collecting $$. Darren