Lack of ZFS File System Support in Snow Leopard Due to Licensing Issues?

[MattInOz]

.....

For a longer, more detailed look at it try http://storagemojo.com/zfs-threat-or-menace-pt-i/
......

Robin Harris
StorageMojo.com

So what your telling us is that Apple have dumped or put on hold ZFS in favour of GFS after all the licensing issues with GFS would seem to be going away if Google are planning it to be part of the OS they want to release. Probably part of Wave as well.

What better joint venture is there. Google bring years of getting the FS to break in nice ways. Apple bring years of giving obscure tech elegant faces that break in nice ways for everyday people. Or better still not break, but we all no thats a pipe dream.

Both bring some think to the table, and some thing of value to gain from the deal.

 

[overcast]

So that is exactly backwards as to who is asking for indemnification. Apple would be asking Sun to cover any awards to NetApp.


Until the dust settles on that, Apple probably doesn't want to put 10-20 million copies of something out there that has a significant chance of loosing in a patent battle. If that turned out to be $10-20 a copy penalty for the violation that is a big chunk of money.

Furthermore, because Sun is in limbo there is probably no one at Sun right now who could sign Sun for such a huge liability. Selling yourself to another company and that taking on a huge liability in the middle of the long transaction usually leads to trouble (get sued and/or costs more money) or at the very least political suicide (tagged as someone who saddles company with risks... not going to make the "keeper" list on the merger process.) .

Even if was in the direction you stated. Sun wanted indemnification for changes Apple made to ZFS to jam it into OS X that for some reason Sun was going to merge back into the core ZFS code tree.


There are also technical issues you are sweeping under the rug.

ZFS changes the why the kernel goes about dealing with the "file system". It isn't a file system. It is a volume manager merged with a file system. That is anti-Linux design philosophy. No way Linus would vote thumbs up on this even if it had GPL license on it.

[ Besides Oracle "liked" Linux in part because it didn't have an operating system. Now that they have Solaris ... it is dubious to "give away" one of the crown jewels of Solaris , ZFS, to Linux. Putting people on Solaris is a better opportunity for Oracle to make money than on Linux. In selling support contracts for OpenSolaris ( or the non open version) Oracle is the #1 player. In Linux, they are much smaller player to Redhat.

Besides when has Oracle taken a software product and slapped a GPL licesnse on it. There was stuff that was already GPL. Or defacto had to be GPL'd (core Linux additions). But name something that Oracle has injected into the GPL space when didn't "have to" ?

The GPL was/is useful in reducing the price of largely, historically, complementary offerings. If people pay less for OS and server hardware can pay MORE for Oracle stuff. Reduction of cost of complements.

Now that they will be in the OS and hardware game. Why accelerate knocking down complements?

Even more so in that Solaris/ZFS is a critical enabler of Suns universal storage devices. Those bring in
cold hard cash money. When the last time you saw Oracle walk away from cash money?????????????
]



Likewise squeezing this into Mac OS kernel would have issues both in user experience ( people just unplugging disks ) and in weaving those mods in with all of the other Snow Leopard updates in flight.

ZFS made much more sense when Apple still had XRaid and might have been inclined to expand past just 1U servers. Backtracking on XRaid. Having XSan , Time Machine success (in the consumer space), and barely treading water on servers ..... the biggest 'bang for the buck" for ZFS is not there; especially in the short term.

Except for ZFS snapshots, which would take TimeMachine to the level where it should have been the whole time.

 

[deconstruct60]

That's backwards. ZFS makes less sense with XRaid. ZFS is designed for JBODs, not hardware raids.

1. XRaid was a higher end storage device. Now they have none. Wanting to even be in the storage device business is a necessary precursor.

2. XRaid could present as a JBOD. A very expensive JBOD, but a JBOD none the less. Minor hardware change to make it an JBOD only device.
However, that completely misses the point. Could also make a slightly different hardware change ( move a better more general CPU to the XRaid and to 10/1 GB ethernet or SAS as network ) and you would still have a storage server but more along the lines of a Sun Unified Storage solutions.

Myopically, solely viewed as a SAN (multiclient), Fiber Channel (FC) storage server, .... sure ZFS doesn't even play in that space. It is not a network file system ( it is a direct attached storage solution). However your average network storage box does have lots of attached storage. In that context it makes lots of sense.


There were two primary deployments for XServe RAID

a.

[ cloud of clients ] --> Xserve (or generic server ) --FC--> XRaid ( 1 or more)

b.
[ cloud of clients ] -- FC --> XRaid


In the first scenario ZFS absolutely can bring benefits if want to deliver a more cost effective solution. (by chucking FC and a few minor tweaks. Take the RAID controllers out or merging the server/storage box. ) The second one is more a blueprint for XSan. ZFS doesn't displace that. Drop the FC link and make the storage serve an iSCSI target it can. ( Apple would have to stop ignoring iSCSI , higher GB Ethernet , and Inifinband though. )




3. Selling more than one storage product an be economies of scale to bring both products down ( the FC SAN box ) and the more generally useful network storage box (which also drives down the shared components parts with XServe it would have in common since fusing a Xserve and XRaid into a single box. ).

Long term having a FC only solution was folly. FC was not XRaid's most important property IMHO. SAS (in direct attached) , Higher speed Ethernet and Infiniband are displacing FC at either lower cost or better price/performance. Instead the big storage product is time capsule. (which also is a direct attached storage device and not a SAN. ) If Apple is heading the "time capsule" route ZFS makes more sense in the long term (when you embedded solution had multiple cores and decent amount of RAM to work with. ) .

I didn't mean having XRaid solely. Meant XRaid plus siblings to flush out the offering with. If were only going to have a single, over $500, storage product then they did the right thing in killing it off.

 

[deconstruct60]

So what your telling us is that Apple have dumped or put on hold ZFS in favour of GFS

No. That link is to show why he thinks ZFS is a good thing. What the heck does the Google File system have to do with any single computer file system????

In so far as GFS is a distributed file system and ZFS isn't, the comparison isn't on substitutability or equivalence but on robustness.

 

[deconstruct60]

Marketers often ask the question "would you prefer 100% of nothing or 40% of something huge?"

Stupid question here, if ZFS is GPLed and Apple is able to use it freely, what is Sun supposed to be getting 40% of?

You are applying the percentages to the wrong side. Also nothing to do with GPL (since the ZFS is CDDL already anyway.) [at least I hope ... seriously misguided if not. ]

If ZFS is kept unique to Solaris then Sun/Oracle gets about 100% of the benefit of those using it because can only get to ZFS through them. If they let other folks use ZFS then may make money off of the 40% share of the much larger market that picks the ZFS enabled solutions.

An analogy using Apple retail market would be if Apple only sold macs at their stores (thus get 100% of retail sales) versus letting other retailers sell them at their's in addition to the Apple stores. A much broader market in the latter case. Apple doesn't sell all of the macs though (so smaller share of a bigger pie) .



There is huge fallacy here that if Sun would only release ZFS as GPL that it would instantly jump to the front of the pack for the next primary, up and coming Linux file system. ( that is the SOLE reasonable rationale to re-release it as GPL as opposed to what it currently is) Being attached to Linux will bring great growth and exposure. That is extremely doubtful assertion. ext4 and btrfs already have a head start. ZFS will but heads with not just the filesystem folks in Linux but in the volume management folks. There is no evolution path from ext3/4 to btfs.



Put aside the notion that Larry spent about billion (or at least 100's of millions ) bucks to get Solaris only to kill it off and put the juicey bits into Linux for free. ROTFLMAO. it makes no sense. From Sun/Oracle's perspective ZFS is no more "free" under CDDL as it would be under GPL in any more significant sense. The CDDL being incompatible with GPL is a feature for some. And it is not for others. Has little to do with the breadth of the market ZFS could reach.

 

[MattInOz]

No. That link is to show why he thinks ZFS is a good thing. What the heck does the Google File system have to do with any single computer file system????

In so far as GFS is a distributed file system and ZFS isn't, the comparison isn't on substitutability or equivalence but on robustness.

Who has a single computer system anymore?

Between home, work, phone and media centre. Times any of those by each person in the house, office, school. We duck in and out of the networks these are on all day yet want everything co-ordinated.

If a distributed file system can manage all that for us and give the useful benefits of a pooled system like ZFS then why not look at it.