Hi There, I looked at the ZFS admin guide in attempt to find a way to leverage ZFS capabilities (storage pool, mirroring, dynamic stripping, etc.) for Xen domU file systems that are not ZFS. Couldn''t find an answer whether ZFS could be used only as a "regular" volume manager to create logical volumes for UFS or even a Linux ext2fs, with ideally, the ability to create snapshots and clones... Also, what can we expect, performance wise, from a file-backed logical block device created on ZFS? Generally, performance is bad due to the beneath file system overhead in comparison to a physical block device. Would ZFS perform good enough in this configuration? Thanks - Patrick
On Tue, Aug 01, 2006 at 01:10:44PM +0200, Patrick Petit wrote:> Hi There, > > I looked at the ZFS admin guide in attempt to find a way to leverage ZFS > capabilities (storage pool, mirroring, dynamic stripping, etc.) for Xen > domU file systems that are not ZFS. Couldn''t find an answer whether ZFS > could be used only as a "regular" volume manager to create logical > volumes for UFS or even a Linux ext2fs, with ideally, the ability to > create snapshots and clones...You can create a zvol, which is a ZFS pseudo-device: # zfs create -V 50G pool/myvol This will create a 50G volume at ''/dev/zvol/dsk/pool/myvol''. You can take advantage of standard ZFS features, including compression, snapshots, clones, etc. Of course be careful snapshotting live volumes, you may catch the filesystem in an inconsistent state and end up unhappy if you rollback, clone, etc.> Also, what can we expect, performance wise, from a file-backed logical > block device created on ZFS? Generally, performance is bad due to the > beneath file system overhead in comparison to a physical block device. > Would ZFS perform good enough in this configuration?Do you mean a pool created from files, or a zvol as described above? Creating pools from files is definitely not a good idea if you care about performance. You''ll be going through two layers of filesystems, with all the added overhead of the standard POSIX interface. If you mean the latter, you should get good performance, though it will depend on your workload. The volume talks directly to the DMU (data management unit), bypassing the POSIX layer. You can even tune the volume block size (with -b) to your workload - large streaming reads/writes can benefit from a larger blocksize, depending on whether the filesystem above can really take advantage of it. Hope that helps, - Eric -- Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock
Eric Schrock wrote:>On Tue, Aug 01, 2006 at 01:10:44PM +0200, Patrick Petit wrote: > > >>Hi There, >> >>I looked at the ZFS admin guide in attempt to find a way to leverage ZFS >>capabilities (storage pool, mirroring, dynamic stripping, etc.) for Xen >>domU file systems that are not ZFS. Couldn''t find an answer whether ZFS >>could be used only as a "regular" volume manager to create logical >>volumes for UFS or even a Linux ext2fs, with ideally, the ability to >>create snapshots and clones... >> >> > >You can create a zvol, which is a ZFS pseudo-device: > >That''s great ! Is this documented in the admin guide? I haven''t seen it.> # zfs create -V 50G pool/myvol > >This will create a 50G volume at ''/dev/zvol/dsk/pool/myvol''. You can >take advantage of standard ZFS features, including compression, >snapshots, clones, etc. Of course be careful snapshotting live volumes, >you may catch the filesystem in an inconsistent state and end up unhappy >if you rollback, clone, etc. > >Sure, I understand that.> > >>Also, what can we expect, performance wise, from a file-backed logical >>block device created on ZFS? Generally, performance is bad due to the >>beneath file system overhead in comparison to a physical block device. >>Would ZFS perform good enough in this configuration? >> >> > >Do you mean a pool created from files, or a zvol as described above? > >I meant the former.>Creating pools from files is definitely not a good idea if you careabout performance. You''ll be going through two layers of filesystems,with all the added overhead of the standard POSIX interface. If you mean the latter, you should get good performance, though it will depend >on your workload. The volume talks directly to the DMU (data management >unit), bypassing the POSIX layer. You can even tune the volume block >size (with -b) to your workload - large streaming reads/writes can >benefit from a larger blocksize, depending on whether the filesystem >above can really take advantage of it. > >Hope that helps, > >Yes it does completely. Thank you. - Patrick>- Eric > >-- >Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock > >-- Patrick Petit Sun Microsystems Inc. Labs, CTO - G2 Systems Exp. ICNC Grenoble (http://icncweb.france) Phone: (+33)476 188 232 x38232 180, Avenue de l''Europe Fax: (+33)476 188 282 38334 Saint-Ismier Cedex, France
Hi Patrick, Here''s a pointer to the volume section in the ZFS admin guide: http://docsview.sfbay/app/docs/doc/817-2271/6mhupg6gl?a=view I welcome any comments-- Cindy Patrick Petit wrote:> Eric Schrock wrote: > >> On Tue, Aug 01, 2006 at 01:10:44PM +0200, Patrick Petit wrote: >> >> >>> Hi There, >>> >>> I looked at the ZFS admin guide in attempt to find a way to leverage >>> ZFS capabilities (storage pool, mirroring, dynamic stripping, etc.) >>> for Xen domU file systems that are not ZFS. Couldn''t find an answer >>> whether ZFS could be used only as a "regular" volume manager to >>> create logical volumes for UFS or even a Linux ext2fs, with ideally, >>> the ability to create snapshots and clones... >>> >> >> >> You can create a zvol, which is a ZFS pseudo-device: >> >> > That''s great ! Is this documented in the admin guide? I haven''t seen it. > >> # zfs create -V 50G pool/myvol >> >> This will create a 50G volume at ''/dev/zvol/dsk/pool/myvol''. You can >> take advantage of standard ZFS features, including compression, >> snapshots, clones, etc. Of course be careful snapshotting live volumes, >> you may catch the filesystem in an inconsistent state and end up unhappy >> if you rollback, clone, etc. >> >> > Sure, I understand that. > >> >> >>> Also, what can we expect, performance wise, from a file-backed >>> logical block device created on ZFS? Generally, performance is bad >>> due to the beneath file system overhead in comparison to a physical >>> block device. Would ZFS perform good enough in this configuration? >>> >> >> >> Do you mean a pool created from files, or a zvol as described above? >> >> > I meant the former. > >> Creating pools from files is definitely not a good idea if you >> careabout performance. You''ll be going through two layers of >> filesystems,with all the added overhead of the standard POSIX >> interface. If you mean the latter, you should get good performance, >> though it will depend >> on your workload. The volume talks directly to the DMU (data management >> unit), bypassing the POSIX layer. You can even tune the volume block >> size (with -b) to your workload - large streaming reads/writes can >> benefit from a larger blocksize, depending on whether the filesystem >> above can really take advantage of it. >> >> Hope that helps, >> >> > Yes it does completely. Thank you. > > - Patrick > >> - Eric >> >> -- >> Eric Schrock, Solaris Kernel Development >> http://blogs.sun.com/eschrock >> >> > >
Sorry, here''s the correct URL: http://docs.sun.com/app/docs/doc/817-2271/6mhupg6gl?a=view Cindy Al Hopper wrote:> On Tue, 1 Aug 2006, Cindy Swearingen wrote: > > >>Hi Patrick, >> >>Here''s a pointer to the volume section in the ZFS admin guide: >> >>http://docsview.sfbay/app/docs/doc/817-2271/6mhupg6gl?a=view > > ^^^^^^^^^^^^^^ > > Hi Cindy, > > That URL is inaccessible outside of Sun. > > Regards, > > Al Hopper Logical Approach Inc, Plano, TX. al at logical-approach.com > Voice: 972.379.2133 Fax: 972.379.2134 Timezone: US CDT > OpenSolaris.Org Community Advisory Board (CAB) Member - Apr 2005 > OpenSolaris Governing Board (OGB) Member - Feb 2006
For those outside of Sun you can get the same at http://docs.sun.com/app/docs/doc/817-2271/6mhupg6gl?a=view -Angelo On 1 Aug 2006, at 11:37, Cindy Swearingen wrote:> Hi Patrick, > > Here''s a pointer to the volume section in the ZFS admin guide: > > http://docsview.sfbay/app/docs/doc/817-2271/6mhupg6gl?a=view > > I welcome any comments-- > > Cindy > > > Patrick Petit wrote: >> Eric Schrock wrote: >>> On Tue, Aug 01, 2006 at 01:10:44PM +0200, Patrick Petit wrote: >>> >>>> Hi There, >>>> >>>> I looked at the ZFS admin guide in attempt to find a way to >>>> leverage ZFS capabilities (storage pool, mirroring, dynamic >>>> stripping, etc.) for Xen domU file systems that are not ZFS. >>>> Couldn''t find an answer whether ZFS could be used only as a >>>> "regular" volume manager to create logical volumes for UFS or even >>>> a Linux ext2fs, with ideally, the ability to create snapshots and >>>> clones... >>>> >>> >>> >>> You can create a zvol, which is a ZFS pseudo-device: >>> >> That''s great ! Is this documented in the admin guide? I haven''t seen >> it. >>> # zfs create -V 50G pool/myvol >>> This will create a 50G volume at ''/dev/zvol/dsk/pool/myvol''. >>> You can >>> take advantage of standard ZFS features, including compression, >>> snapshots, clones, etc. Of course be careful snapshotting live >>> volumes, >>> you may catch the filesystem in an inconsistent state and end up >>> unhappy >>> if you rollback, clone, etc. >>> >> Sure, I understand that. >>> >>>> Also, what can we expect, performance wise, from a file-backed >>>> logical block device created on ZFS? Generally, performance is bad >>>> due to the beneath file system overhead in comparison to a >>>> physical block device. Would ZFS perform good enough in this >>>> configuration? >>>> >>> >>> >>> Do you mean a pool created from files, or a zvol as described above? >>> >> I meant the former. >>> Creating pools from files is definitely not a good idea if you >>> careabout performance. You''ll be going through two layers of >>> filesystems,with all the added overhead of the standard POSIX >>> interface. If you mean the latter, you should get good performance, >>> though it will depend >>> on your workload. The volume talks directly to the DMU (data >>> management >>> unit), bypassing the POSIX layer. You can even tune the volume block >>> size (with -b) to your workload - large streaming reads/writes can >>> benefit from a larger blocksize, depending on whether the filesystem >>> above can really take advantage of it. >>> >>> Hope that helps, >>> >> Yes it does completely. Thank you. >> - Patrick >>> - Eric >>> >>> -- >>> Eric Schrock, Solaris Kernel Development >>> http://blogs.sun.com/eschrock >>> > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss