Linked by Thom Holwerda on Mon 15th Nov 2010 23:37 UTC, submitted by comay
Oracle and SUN Today Oracle released its latest version of Solaris technology, the Oracle Solaris 11 Express 2010.11 release. It includes a large number of new features not found in either Oracle Solaris 10 or previous OpenSolaris releases including ZFS encryption and deduplication, network-based packaging and provisioning systems, network virtualization, optimized I/O for NUMA platforms and optimized platform support including support for Intel's latest Nehalem and SPARC T3. In addition, Oracle Solaris 10 support is available from within a container/zone so migration of existing systems is greatly simplified. The release is available under a variety of licenses including a supported commercial license on a wide variety of x86 and SPARC platforms.
Thread beginning with comment 450128
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[2]: ZFS
by kaiwai on Tue 16th Nov 2010 13:26 UTC in reply to "RE: ZFS"
kaiwai
Member since:
2005-07-06

ZFS was originaly touted on Apple website for one of previous releases of OSX (10.5 or 10.6 can't remember) Of course it didn't help that Jonathan Schwartz had to open his big mouth and announce that it was gonna be default filesystem in OSX. Jobs just doesn't like other people pre-announcing stuff. Given the rapprochement between Oracle/Apple over Java we might be lucky and see ZFS as part of 10.7 (lion).

That's more of a pipedream on my part but you never know. The recent "back to Mac" event said very little about Lion other then about some of the GUI stuff and the release name. Given that OSX currently uses DTrace and Solaris NFSv4 stack (client/server) I wouldn't be surprised if ZFS snuck back in. Helps that the NetApp case has ended. Plus ZFS is even better now then it was two years ago with stuff like DeDup and Crypto now included.


From what I understood it all came down to patents and licensing plus other stuff - stuff that could have over come if Sun (and subsequently Oracle) had the will power to do so. Even if they did get ZFS working with Mac OS X the problem I found when using it with OpenSolaris was that the performance was terrible on anything less than 2GB of RAM using a 32bit kernel on a low powered machine - which would have pretty much killed off using it on the MacBook Air. What Apple needs isn't something revolutionary but rather something that is evolutionary and does what is required without too much fanfare.

One possible replacement for HFS+ that comes to mind could be HAMMERFS from DragonflyBSD:

http://en.wikipedia.org/wiki/HAMMER

It includes many cool features and is licensed under a liberal license which will allow Apple to bring it over to XNU and make the necessary enhancements for Finder integration. Sure, it isn't a massive leap in the case of ZFS but the file system sticks to the tried and true way of doing things - I'd sooner have that than risk a jump into the unknown for the sake of having something on the bleeding edge.

Reply Parent Score: 2

RE[3]: ZFS
by Kebabbert on Tue 16th Nov 2010 14:22 in reply to "RE[2]: ZFS"
Kebabbert Member since:
2007-07-27

the problem I found when using ZFS with OpenSolaris was that the performance was terrible on anything less than 2GB of RAM using a 32bit kernel on a low powered machine

The problem was the 32bit cpu. ZFS is 128 bit, and does not like 32 bit cpus, performance will suck. I used 32bit Pentium4 and 1GB RAM for a year, and I got 20-30MB/sec with a ZFS raid with 4 discs. This sucks indeed.

If you use 64 bit cpus, then performance will be normal.

The reason ZFS gets slow, is because ZFS does lots of checksumming and computations with respect to data integrity. ZFS wants to guarantee that your data is not subject to bit rot or silent corruption. Other filesystems or solutions, dont do that. Why do you think they are much faster than ZFS? Because they dont do the expensive calculations with respect to data safety!

Reply Parent Score: 2

RE[4]: ZFS
by kaiwai on Tue 16th Nov 2010 17:30 in reply to "RE[3]: ZFS"
kaiwai Member since:
2005-07-06

The problem was the 32bit cpu. ZFS is 128 bit, and does not like 32 bit cpus, performance will suck. I used 32bit Pentium4 and 1GB RAM for a year, and I got 20-30MB/sec with a ZFS raid with 4 discs. This sucks indeed.

If you use 64 bit cpus, then performance will be normal.


But on low powered 64bit processors such as the MacBook Air 1.4Ghz the experience isn't exactly going to be all that pleasant to say the least. The amount of memory is a big killer - give it 4GB of memory and it flies but I say 2GB minimum but even then it isn't all too pleasant. It is a great file system designed for a large system with a tonne of memory but for something that is low powered requiring something light weight it is probably not the ideal file system to use.

The reason ZFS gets slow, is because ZFS does lots of checksumming and computations with respect to data integrity. ZFS wants to guarantee that your data is not subject to bit rot or silent corruption. Other filesystems or solutions, dont do that. Why do you think they are much faster than ZFS? Because they dont do the expensive calculations with respect to data safety!


People must have some really bad luck because of all the things that have pulled down my computer I've never experienced file system corruption - 9/10 if something goes haywire it is because of my own doing rather than an act of the computer gods trying to smite me with a kernel panic and file system corruption.

In the 10 years I've owned Mac's (from an eMac to the current iMac and MacBook Pro) I haven't experienced a single case of file system corruption *touch wood*

Reply Parent Score: 2

RE[3]: ZFS
by Laurence on Tue 16th Nov 2010 14:57 in reply to "RE[2]: ZFS"
Laurence Member since:
2007-03-26

What Apple needs isn't something revolutionary but rather something that is evolutionary and does what is required without too much fanfare. One possible replacement for HFS+ that comes to mind could be HAMMERFS from DragonflyBSD: http://en.wikipedia.org/wiki/HAMMER It includes many cool features and is licensed under a liberal license which will allow Apple to bring it over to XNU and make the necessary enhancements for Finder integration. Sure, it isn't a massive leap in the case of ZFS but the file system sticks to the tried and true way of doing things - I'd sooner have that than risk a jump into the unknown for the sake of having something on the bleeding edge.


I've not read up on HAMMERFS yet (will read your wiki shortly), but ZFS would have been perfect for the media professional market. Many OS X users are photographers, musicians, music producers or movie producers. They deal with huge files, many of which are duplicates of each other minus small alterations. So ZFSs deduping and/or the intelligent copy feature (where only the differences are stored when a file is copied).

Reply Parent Score: 2

RE[4]: ZFS
by kaiwai on Tue 16th Nov 2010 17:36 in reply to "RE[3]: ZFS"
kaiwai Member since:
2005-07-06

I've not read up on HAMMERFS yet (will read your wiki shortly), but ZFS would have been perfect for the media professional market. Many OS X users are photographers, musicians, music producers or movie producers. They deal with huge files, many of which are duplicates of each other minus small alterations. So ZFSs deduping and/or the intelligent copy feature (where only the differences are stored when a file is copied).


Actually data de-dup is one of the features of HAMMERFS ;)

http://en.wikipedia.org/wiki/HAMMER

* Keeps configurable fine-grained filesystem history (30-second boundaries typically).
* Keeps configurable coarse-grained filesytem history (60 days daily by default).
* History and snapshots are accessible via the live filesystem.
* Near real-time log-less streaming mirroring to slaves or backups.
* Slave can have different retention parameters.
* CRC checksumming of data and metadata.
* Minimal remount time (no fsck required).
* Designed from the ground up for large fan-out mirroring.
* Support for very large file-systems (up to 1 exabyte).
* Offline recovery after a catastrophic failure is possible.
* Ability to re-optimize the layout in the background, on a live filesystem.
* Data de-dup
* Future - multi-master clustering
* Future - data compression


I've given DragonflyBSD a tinker and what I notice is there isn't the same sort of performance penalty as I notice with ZFS so that alone makes me a big booster for the file system; first impressions are sure to win me over and HAMMERFS has definitely done that ;)

Reply Parent Score: 2