Date: Thu, 28 Apr 2011 20:45:33 +0000 From: Malcolm Waltz <mwaltz@PACIFIC.EDU> To: Chuck Swiger <cswiger@mac.com> Cc: FreeBSD Stable <freebsd-stable@freebsd.org> Subject: Re: ZFS vs OSX Time Machine Message-ID: <D4F10CE3-C049-4CEE-A52B-AECF3C12BEA7@pacific.edu> In-Reply-To: <8D2285F1-3706-4FEB-A4B4-10089AC7A622@mac.com> References: <537A8F4F-A302-40F9-92DF-403388D99B4B@gsoft.com.au> <2B80846C-E8A9-4FF6-962C-9405469661D6@mac.com> <BANLkTimdhwtj2q=jEC_dTU7Brv7g6mHMUQ@mail.gmail.com> <8D2285F1-3706-4FEB-A4B4-10089AC7A622@mac.com>
next in thread | previous in thread | raw e-mail | index | archive | help
AFP is not the same as HFS+. Time Machine will work better with AFP than N= FS or SMB/CIFS, but it's still not using native HFS+ unless you are using b= lock storage (even if you use AFP with an HFS+ filesystem). Time Machine cannot function at all without accessing HFS+ directly. If yo= u are using a network filesystem (AFP, SMB/CIFS or NFS), Time Machine creat= es a sparse disk image, formatted as HFS+ and stores it on your file-server= . It then attaches that disk image as a disk device and mounts it (somewha= t like "mdconfig -a -t vnode -f /path/to/disk-image -u 1; mount /dev/md1 /m= nt"). It then treats that disk image basically the same way that it treats= local attached storage, including creating hard directory links (but all i= nside the disk image). See man hdiutil (on OS X) for more info, particular= ly the part about SPARSEBUNDLEs, sparse images backing HFS+ filesystems and= band sizes. Even if you use Mac OS 10 Server and create a Time Machine share (which is = the best case scenario), it still uses emulated block storage as described = above (disk image over AFP on HFS+). I have personally done this and decid= ed that it was not a very good solution. Your milage may very. I know tha= t people do this, but it seems rather silly. If you have the knowledge to use ZFS, use a zvol via iSCSI. It is much mor= e efficient to use a form of network storage that handles block access nati= vely (like iSCSI) instead of accessing emulated block storage via file-shar= ing protocols that were not designed for such use. ZFS doesn't care what y= ou use it for. If you are using ZFSv28 (I wouldn't use it for critical dat= a on FreeBSD yet) you can even do dedupe and compression on a native HFS+ T= ime Machine volume (although you would only see the saved space from the pe= rspective of the zpool and make sure you have lots of RAM).=20 On Apr 28, 2011, at 12:33 PM, Chuck Swiger wrote: > On Apr 28, 2011, at 12:17 PM, George Kontostanos wrote: >> I am using TM over smb on a ZFS Raidz1 pool of my fileserver with no pro= blems whatsoever. =20 >>=20 >> NAME USED AVAIL REFER MOUNTPOINT >> tank/apple 37.2G 82.8G 37.2G /tank/apple >>=20 >> Oldest backup 14 December 2009 >=20 > SMB aka CIFS is a better choice than NFS, because it supports better lock= ing (oplocks or "stealable" locks), but it is not as good as AFP for this p= articular purpose. Also, ZFS isn't going to be as space efficient at stori= ng TM backups compared with HFS+, because it doesn't support hard links to = directories. >=20 > Regards, > --=20 > -Chuck >=20 >=20 >=20 > _______________________________________________ > freebsd-stable@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-stable > To unsubscribe, send any mail to "freebsd-stable-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?D4F10CE3-C049-4CEE-A52B-AECF3C12BEA7>