Forgot your password?
typodupeerror
Businesses Data Storage Open Source Patents Sun Microsystems The Courts Hardware Your Rights Online

NetApp Threatens Sellers of Appliances Running ZFS 231

Posted by Soulskill
from the they-also-invented-the-box-and-the-bucket dept.
eldavojohn writes "ZFS is licensed under the CDDL and is considered to be open source, but NetApp is sending threatening legal letters to startups who look to offer ZFS on NAS appliances. This assault on Coraid has a few people worried about the future of ZFS as NetApp rears its ugly head yet again. The CEO of Coraid replied to NetApp's demands, saying, 'We made the decision to suspend shipment after receiving a legal threat letter from NetApp Inc., suggesting that the open-source ZFS file system planned for inclusion with our EtherDrive Z-Series infringes NetApp patents.' Will NetApp effectively destroy any future ZFS might have enjoyed?"
This discussion has been archived. No new comments can be posted.

NetApp Threatens Sellers of Appliances Running ZFS

Comments Filter:
  • why? (Score:3, Insightful)

    by pak9rabid (1011935) on Friday July 09, 2010 @02:57PM (#32853498)
    Wait...doesn't the filesystem get created by the user of the storage device? I.E., the storage system gets hooked up to your system(s), the OS sees them as raw block devices, then the user generally puts the filesystem on top of the block device.....is this not the way the EtherDrives work?
    • Re:why? (Score:4, Informative)

      by natehoy (1608657) on Friday July 09, 2010 @03:02PM (#32853574) Journal

      What you say is true for LOCAL storage, but EtherDrives are NAS (Network Attached Storage) as mentioned in the summary.

      This means they come preformatted, but the machines that access the storage are using Samba or Windows File Sharing or whatever to access it, so the client PCs do not see the filesystem on the NAS box.

      • This means they come preformatted

        Solution: make it format itself the first time the end user turns it on.

      • OK...seeing as how Linux doesn't have kernel support for ZFS, who were they planning on marketing this to, the Solaris and BSD shops? Seems like a bad idea to sell a NAS device that isn't compatible with the most popular server OS.
        • Re: (Score:3, Insightful)

          by pak9rabid (1011935)
          Ugh..nevermind..I misread that...me = stupid.
        • Re:why? (Score:5, Informative)

          by natehoy (1608657) on Friday July 09, 2010 @03:44PM (#32854176) Journal

          I know you've already replied that you misunderstood, but just in case anyone else is confused by this...

          This is a NAS, which is itself a server. Support for the filesystem is built into the NAS box. The NAS box then exposes the data it stores on that filesystem to the network using network-appropriate protocols.

          Anyone wanting to access it would use a networking standard like Samba, Windows File Sharing, FTP, or whatever services the NAS box allows.

          Of course, they'd also access the management tools (nowadays generally a small web server also built into the NAS box).

          None of the clients would need to support the underlying filesystem that the NAS box uses. In fact, they wouldn't even be allowed to know what that filesystem is.

          Back when I had Windows boxes at home, they had absolutely no problems reading shares I made on my Linux box. The Linux box could be formatted ext, Reiser, or anything I wanted that Linux supported.

          As long as I never tried to take a hard drive out of the Linux box and put it in the Windows box, of course. Then it becomes local storage, and Windows would have to support the filesystem in order to read it.

    • NetApp (Score:4, Interesting)

      by goombah99 (560566) on Friday July 09, 2010 @03:21PM (#32853842)

      Much as I hate these patent cases, perhaps this one has merit. NetApp built it's bussiness being a vendor of NAS systems that had extensible file systems that spanned clever raid structures, and automatic snapshoting and they did this long before ZFS. Those are the key features of ZFS. And when you pair that with NAS, well that's a NetApp in a box. I dont know what NetApps patents claim but what they did was not obvious at the time and they are actively a seller of that, not a patent troll.

      • Re:NetApp (Score:5, Insightful)

        by Just Some Guy (3352) <kirk+slashdot@strauser.com> on Friday July 09, 2010 @03:36PM (#32854066) Homepage Journal

        NetApp built it's bussiness being a vendor of NAS systems that had extensible file systems that spanned clever raid structures, and automatic snapshoting and they did this long before ZFS. Those are the key features of ZFS.

        VMS did it earlier. Screw NetApp and their overpriced, underfeatured, patented crap. Really. I mean that.

        • Here, here... (Score:5, Insightful)

          by multimediavt (965608) on Friday July 09, 2010 @04:09PM (#32854474)

          Screw NetApp and their overpriced, underfeatured, patented crap. Really. I mean that.

          I totally and completely agree with that summary of NetApp

          With that out of the way, how does NetApp have any authority to enforce a license/patent on a piece of software they did not invent, nor hold the licensing for? ZFS was created by Sun and released under the CDDL. I am confused as to where NetApp fits into this equation other than being a troll of something that isn't even theirs to begin to troll with. I will do some digging online, but this is just effed up.

          • Again... (Score:4, Insightful)

            by Junta (36770) on Friday July 09, 2010 @04:29PM (#32854742)

            NetApp is suing Sun (and now Oracle) over ZFS because they *claim* it violates patents they hold that they implemented in their own WAFL fs. WAFL does predate ZFS. NetApp was granted patents in WAFL that ZFS seems to also do, but prior-art may have caused those patents to be invalid.

            They are suing vendors that also use ZFS probably because those vendors are either licensing from Oracle (in which case indemnification may just pass this on) or are using FreeBSD, which makes the link between the vendor and Oracle a bit more odd and possible has the vendor on the hook for liability.

          • Re:Here, here... (Score:5, Interesting)

            by mea37 (1201159) on Friday July 09, 2010 @05:16PM (#32855334)

            Patents are funny that way. If I patent an invention and use it in Product X, and you come along and create Product Y which infringes my patent, then I can sue you (as you expect), and also I can sue anyone who uses, sells, or offers for sale Product Y. That's one of the reasons indemnity is a big deal when corporations look at OSS solutions, and it was the basis of many of SCO's legal threats back in the day.

            So you get the store down the street to sell Product Y, I can sue the store down the street. Bob buys Product Y from the store and puts it to use, in theory I can probably sue Bob. Sue enters into an OEM arrangement with you and embeds Product Y in Product Z, I can sue Sue, and Sue's distributors and users as well. All because I have a patent on something I did in Product X.

            • Re: (Score:3, Insightful)

              by sjames (1099)

              Sadly, all of that is true even if product Y was invented entirely from first principles by people with no knowledge of X, even if they did it at the same time as X was being invented.

              How anyone can claim that granting the patent to inventor X doesn't steal from inventor Y the fruits of his labor, I cannot say.

          • Re: (Score:3, Interesting)

            by afidel (530433)
            Netapp is claiming that ZFS infringes on multiple core WAFL (the filesystem behind their filer products) patents. They may have a point insofar as software patents are enforceable. Time and again the Sun engineers behind ZFS have referenced WAFL as being the closest comparable filesystem to ZFS. The patent enforcement against ZFS started after Sun in their pre-Oracle takeover death throws came to Netapp expecting $$$ millions for supposed patent infringement.
          • Re:Here, here... (Score:5, Interesting)

            by Luminary Crush (109477) on Friday July 09, 2010 @06:17PM (#32855928)

            So the story goes, this dates back to some interns who worked at NetApp and then went to Sun and perhaps influenced ZFS.

            The technology in WAFL is that of a pointer-based filesystem - which itself is pretty ingenious and is only now being feature-emulated (ZFS, BRTFS, etc).

            One can say what they want of Netapp's pricing, but the technology is extremely solid and simple to operate compared to managing Linux or Solaris boxes running a filesystem as a NAS; the snapshots are without I/O penalty and you can take a lot of them, the clustering is *FAR* simpler than anything happening on general-purpose OSes, the support for protocols is industry-leading (FCoE, NFSv4, SMB 2.0 - they have a codeshare w/ Microsoft and do not use a reverse-engineered Samba implementation or run any kind of Windows storage server like competitors do).

            ZFS has a lot of promise, but does not have nearly the performance that WAFL does (considering RAID-DP versus ZFS RAID6) and has only some of the feature set of mirroring, snapshot vaulting, filesystem and file cloning, WORM-compliance, etc. Companies don't want to bet their business on a science project of roll-your-own NAS which doesn't have the feature set the Netapps do, and no serious competitor (eg a company with the ability to financially stand behind the product) in the enterprise space has anything like the feature set.

            I work for a systems integrator and I've messed with hundreds of Netapps, Sun and Linux appliances, and competitors over the years. I use ZFS at home because I can't afford a Netapp (and wouldn't want to pay the electricity bill if I could!) but if I ran an IT department I'd put my data on a Netapp FAS over a ZFS appliance any day.

      • by rayvd (155635)

        I can see what you're saying as well. ZFS does a _lot_ of the things WAFL has done for years. I don't really think those approaches should necessarily be able to be patented though, but as it is now, you can see how something like ZFS -- very similar to WAFL in a lot of ways would raise some red flags and at least warrant some investigation (whether or not we agree with the principles there).

        NetApp should focus on their business model though instead. WAFL is still significantly more mature than ZFS and h

        • Re: (Score:3, Interesting)

          by Just Some Guy (3352)

          ZFS does a _lot_ of the things WAFL has done for years. I don't really think those approaches should necessarily be able to be patented though, but as it is now, you can see how something like ZFS -- very similar to WAFL in a lot of ways would raise some red flags and at least warrant some investigation (whether or not we agree with the principles there).

          Chrome does a _lot_ of the things MSIE has done for years. I don't really think those approaches should necessarily be able to be patented though, but as it is now, you can see how something like Chrome -- very similar to MSIE in a lot of ways would raise some red flags and at least warrant some investigation (whether or not we agree with the principles there).

          I suppose they realize this and are just doing their corporate "due diligence" in aggressively trying to protect their IP.

          Screw their "due diligence". They're just being desperate assholes and everyone knows it.

        • by BitZtream (692029)

          Their stuff "just works".

          You do realize thats pretty much an admission that you've never actually used any of their hardware ... right?

        • So if WAFL is so much better technically then why does NetApp need to throw around bogus patent infringement claims (and yes the claims about ZFS infringing their patents has been ruled to be wrong) rather than just competing on the merit of their product?

          • Re: (Score:3, Insightful)

            by Bert64 (520050)

            Competing on product merit is less profitable because you actually have to invest in improving your product, you can't just keep selling the same old shit at ridiculously high prices and hope the competitors don't overtake and undercut you.

          • WAFL may be better than ZFS, but many customers may decide that it's not $20,000 better.

        • by Bert64 (520050)

          While WAFL may be more mature it won't stay that way...
          NetApp will need to seriously innovate if they want to compete with ZFS based competitors, and that will decrease their profit margins.

      • Re:NetApp (Score:5, Informative)

        by Lunix Nutcase (1092239) on Friday July 09, 2010 @03:58PM (#32854326)

        The case doesn't have merit because the courts have already ruled that ZFS doesn't infringe NetApp's patent No. 6,892,211 which is at the heart of their infringement claims. This is just a shakedown.

      • by BitZtream (692029)

        They may have done it 'before ZFS', but neither of them were original. If they were the first it might matter, but you're not talking about anything new really from them. They were just applying old ideas to new hardware under the hood.

        MS doesn't get a patent for using mouse clicks on the next intel processor because they filed first before Apple any more than NetApp gets something special because they applied 1970s processes to 1990s SCSI drives or SATA/SAS drives now.

        • by BitZtream (692029)

          I feel I should also point out:

          Those are the key features of ZFS. And when you pair that with NAS, well that's a NetApp in a box

          Its also a very obvious pairing to anyone with even a quarter of a clue in the industry, not patent worthy in the least.

      • Re: (Score:3, Interesting)

        by h4rr4r (612664)

        Bullshit, look at mainframes many of those had these features. A NAS is just a server with a bunch of disk.

      • Re: (Score:3, Informative)

        Much as I hate these patent cases, perhaps this one has merit. NetApp built it's bussiness being a vendor of NAS systems that had extensible file systems that spanned clever raid structures, and automatic snapshoting and they did this long before ZFS. Those are the key features of ZFS. And when you pair that with NAS, well that's a NetApp in a box. I dont know what NetApps patents claim but what they did was not obvious at the time and they are actively a seller of that, not a patent troll.

        Sorry, either ZFS violates their patents in all of its uses, or they have no claim. If I build a machine that makes widgets and patent it, then you build a machine that builds widgets using a completely different technique, my patent does not apply to your machine. This sums up a large part of the problem with software patents, it is very hard to see whether a piece of software reaches the same results as another piece of software using different methods or not.

  • NetApp frequently advertises with slashdot.

    NetApp owns the idea of ZFS + NAS? The old "x on the internet" patent attack, eh?

    • NetApp uses WAFL on their NAS. A filesystem they invented before ZFS. They claim ZFS violates patents of WAFL.

      • by poetmatt (793785) on Friday July 09, 2010 @03:35PM (#32854026) Journal

        using WAFL on a nas is not any more inventive than using ZFS on a NAS. Again, the parent is correct.
        Using any filesystem for a NAS is not inventive. It's been around for quite some time. They're also going after distributors rather than attack the ZFS patents they purportedly precede.

        It's also quite impossible to prove any sort of patent violation for using ZFS on a NAS simply due to the competing software being patented.

        If I use a product X, of which is infringing upon product y, I am not liable for uses of product X. The patent system doesn't support 3rd party liability. it's just distributors buckling under legal threats.

        • They are not saying "on a NAS", they are saying ZFS infringes WAFL, regardless of whether it is on a NAS. My impression is that currently some prior art is seeming to render those claims moot, but the crux of it is not it being on a NAS, the crux of it is ZFS as a technology versus WAFL as a technology regardless of context. The comparison of 'x on the internet' isn't quite spot on, it's at least better than that.

          In terms of going after the 'users' rather than the 'developers', the relationship here is ce

        • If I understand correctly, NetApps patents are really about their FS and not NAS, but they had only patented it in NAS context. So using ZFS normally doesn't infringe them, but using it in a NAS does - but the actual infringement is due to algorithms and techniques used in ZFS, and not due to NAS.

        • by mea37 (1201159)

          The validity of the patent can of course be challenged, though my understanding is that "using WAFL with NAS" isn't what's being claimed as patented; some thing that WAFL does is being claimed as patented, and NetApp further claims that ZFS does those same things.

          Also, you seem to think it's legal to sell a product that includes an infringing component; I'm confused why you would think this. Here's an excerpt from 35 USC 271 [uspto.gov], which defines patent infringement:

          (a) Except as otherwise provided in this title,

        • by reebmmm (939463) on Friday July 09, 2010 @05:13PM (#32855296)

          I use a product X, of which is infringing upon product y, I am not liable for uses of product X. The patent system doesn't support 3rd party liability. it's just distributors buckling under legal threats.

          You would be very, very wrong. Your use of an infringing product still makes you an infringer. (35 U.S.C. Sec. 271(a))("whoever without authority makes, uses, offers to sell, or sells any patented invention...). Now, there are default rules that say that a purchaser of a good is "indemnified" from claims of infringement by the seller. (See UCC Art. 2 Sec. 312 [cornell.edu]) This is typically how end user avoid litigation.

          In addition, you're also wrong to think that the patent system doesn't support third party liability. It does. One party's can be sued for another person's direct infringement. It's called "indirect infringement." (35 U.S.C. Sec. 271 (b) and(c)). The rules on when this apply are stated in the statute and are not always so easy to discern. One way to think about it: a party cannot escape liability by simply knowing moving the infringement downstream.

          And, just so that you know, it is a somewhat common strategy that a patent owner sues the end user, direct infringer. There is nothing worse as a supplier than receiving about 30 letters from angry customers seeking indemnification from the infringement suit. This is a great technique to annoy the party with deep pockets and drive home the advantages of settling early.

    • Re: (Score:3, Interesting)

      I suspect that they are going after NAS vendors simply because the NAS vendors are a more direct threat to their business.

      NetApp probably doesn't like that Sun, now Oracle, is using "their" features in direct-attached storage for Solaris boxes; but NetApp doesn't sell direct attached storage, so that isn't as much their problem(and Larry's lawyers are probably scary).

      If, on the other hand, the world starts sprouting outfits who are combining CDDL software+commodity disks+commodity NICs+modestly custom
  • Maybe not NetApp so much as Oracle.
  • Boycott (Score:5, Interesting)

    by Improv (2467) <pgunn@dachte.org> on Friday July 09, 2010 @03:03PM (#32853592) Homepage Journal

    We need a site to organise boycotts of companies that abuse the patent, trademark, or copyright system. Not everyone would need to sign on to all of them, but anyone should be able to post a call and explain their reasoning. If we got enough techies onto it who would use it at work, it could have some muscle.

    • by Grishnakh (216268)

      How many techies actually make purchasing decisions at their workplaces? Not many. Most are made by high-up managers, under advice from sales reps.

      • by h4rr4r (612664)

        I do, if it was that way I would go find a new job in short order. Those folks do not have the knowledge to be making those decisions.

        • by Bert64 (520050)

          Unfortunately, those folks make the purchasing decisions in most organizations...

        • Re: (Score:3, Insightful)

          by Grishnakh (216268)

          Of course they do. They are much better qualified to make purchasing decisions than the low-level people who will be working with the products selected. After all, the sales reps from places like NetApp told them that their expensive products are better, over expensive dinners with wine. Who do you think you are, someone important?

      • Re:Boycott (Score:5, Insightful)

        by sglewis100 (916818) on Friday July 09, 2010 @07:22PM (#32856520)

        How many techies actually make purchasing decisions at their workplaces? Not many. Most are made by high-up managers, under advice from sales reps.

        Where on earth do you work, and can you find better? I'm not currently in a decision making role, although I have been, but I am in a role where I help gather requirements, evaluation specifications, invite vendors for product demos, work on getting evals / visiting labs / visiting and talking with other customers, review quotes, beat up reps on pricing, make a formal recommendation, and wait for my boss to get it done. We have a purchasing department... but they just purchase things they are told to. We also have a CFO, but if we're doing our diligence, and meeting our budget, he's not often overruling us.

    • by Old97 (1341297)
      I have a patent on such sites.
    • Re: (Score:3, Insightful)

      We need a site to organise boycotts of companies that abuse the patent, trademark, or copyright system. Not everyone would need to sign on to all of them, but anyone should be able to post a call and explain their reasoning.

      It's called Slashdot.

      If we got enough techies onto it who would use it at work, it could have some muscle.

      Practice has shown that you won't get enough techies that are actually in charge of software/hardware purchase decisions on any scale sufficient for it to have some muscle.

  • So, what happened? (Score:2, Interesting)

    by somaTh (1154199)
    The Wikipedia article [wikipedia.org] on NetApp talks about them bringing a suit against Sun/ZFS and that Sun countersued, but doesn't mention the verdict. What happened there and why isn't the verdict applying here?
  • We need to announce loudly on Oracle forums that NetApp says that Solaris shops are using illegal software. There are few things that get Larry Ellison moving faster than the idea that someone might cost him a dollar.

    • NetApp has been in a legal battle over WAFL v. ZFS from before the Oracle takeover. Oracle is keenly aware of all of this.

    • At best, oracle would not care. At worst, they would encourage the lawsuit. This says nothing about zfs, only about freedom to use it. Oracle can give customers indemnification on solaris/zfs based stuff. What can Nexenta do, other than provide a great product? Legall, they can only be on the defense, as they probably do not have enough patents for a counter-offense. Oracle can only benefit from this, as nexenta is a competitor.
  • So much for our magical universal file system.

    Another case of cross-platform standards being ruined by a whiny patent abuser.

  • by KonoWatakushi (910213) on Friday July 09, 2010 @03:12PM (#32853704)

    See the second post in the recent thread on zfs-discuss: Legality and the future of zfs... [opensolaris.org]

    It doesn't sound as if Netapp has a leg to stand on, so they are trying to shake down the companies while they can. Where have we seen this before?

  • by bigredradio (631970) on Friday July 09, 2010 @03:15PM (#32853742) Homepage Journal
    I guess this explains OSX Server walking away from ZFS because of "license issues". Since Apple walked away in the 11th hour, I wonder if there is some real validity to NetApp's argument (at least legally).
    • by Joce640k (829181) on Friday July 09, 2010 @03:20PM (#32853816) Homepage

      ZFS isn't *essential* to Apple so why would they fight a patent battle in court? There's plenty of other file systems in the sea.

      • by hedwards (940851)
        Unless it turns out to be an easy suit, I think it would be far more likely that Apple would just pay a licensing fee to use it. I don't think they pulled it over IP concerns, I think it was pulled because it wasn't ready for prime time.
        • by Amouth (879122)

          If it was something that is non essential - and there is a rather long dispute - whom do they "license" it from? do they pay NetApp extortion money? maybe adding some validity to it? or do they just ignore the 10% of people using it and say we are going in this direction with no rime or reason (not uncalled for by Apple) and then after the dispute is over and there is a clear victory - they will add it back in as "fully supported" and the new way to go (cover your ears * whisper* isn't at all like the old

        • Re: (Score:3, Insightful)

          In order to license something, the licensor needs to be willing to enter into an agreement. Don't be so sure that NetApp wants to do that, when they can sell overpriced filers to Apple's customers.

    • by Lunix Nutcase (1092239) on Friday July 09, 2010 @03:23PM (#32853874)

      Except that the courts and the patent office are saying otherwise in NetApp's case against Sun.

    • by eddy (18759) on Friday July 09, 2010 @03:24PM (#32853884) Homepage Journal

      We should all have learned by now that how people and companies react to legal threats doesn't have ANYTHING to do with so-called merits. 'The' SCOX case went on for about seven years and they had NOTHING. People paid them "linux tax" while they had NOTHING, could show NOTHING, made increasingly bizarre and outrageous claims and could in fact never WIN ANYTHING. People still paid.

      The lesson from that of course is that being a tick on the ass of the system it's a perfectly valid way to lift a nice salary and appear important, so really, why not?

      • by ais523 (1172701)
        The SCOX case is still going on. SCO had two appeals left (more if the Supremes decide they want to accept a case from them), and they just used one of them. (I haven't seen what their arguments are, and I don't think it's public knowledge yet; they should be hilarious when they come out, though.)
  • Indemnification (Score:3, Insightful)

    by rayvd (155635) on Friday July 09, 2010 @03:20PM (#32853818) Homepage Journal

    Oracle should offer to provide indemnification to vendors. They've got a large patent portfolio of their own and obviously large assets to make them a much more formidable foe to NetApp.

    • Re:Indemnification (Score:5, Informative)

      by Lunix Nutcase (1092239) on Friday July 09, 2010 @03:25PM (#32853898)

      Sun already does indemnify it's customers. Schwartz pointed this out [sun.com] when NetApp's rumblings against Sun first happened.

      First, the basics. Sun indemnifies all its customers against IP claims like this. That is, we've always protected our markets from trolls, so customers can continue to use ZFS without concern for spurious patent and copyright issues. We stand behind our innovation, and our customers.

      • Re: (Score:3, Interesting)

        by Macrat (638047)

        Sun already does indemnify it's customers. Schwartz pointed this out [sun.com] when NetApp's rumblings against Sun first happened.

        Neither Schwartz or Sun exists anymore.

    • However, Oracle does(or at least Sun did) sell ZFS-based NAS/SAN appliances(pretty cool looking, actually. 4U or so, slides out with top-loading trays for 48 SATA drives. Web interface for carving it up into pools and serving it up over ethernet).

      Thus, while they certainly have some motivation to provide indemnification to customers, because having their server users getting sued is awkward, it isn't clear that they actually mind NetApp shooting their smaller competitors in the ZFS NAS/SAN appliance spac
    • by canajin56 (660655)

      As as been pointed out, they do provide indemnification to vendors and users. However, indemnity doesn't make you immune from lawsuit. Patents, like it or not, protect the use and manufacture of a patented device, not just the distribution of that device. So, Sun, the vendor, and the customer all, separately, violated the parent. The patent holder thus has the right to sue separately all 3 of those entities. And, they can't lose their right to sue the vendor because of what it says in a contract they

  • NetApp must die. And they can be buried in the same grave with NTP.
  • by guruevi (827432) <evi@smo k i n g c ube.be> on Friday July 09, 2010 @03:27PM (#32853912) Homepage

    ZFS is a game changer in the storage industry. While people are buying $250,000 NetApp installations, the exact same hardware, performance and connectivity will go for $5000 of high-end hardware and a couple of hours work with ZFS. $250,000 will easily buy you a Petabyte worth of redundant ZFS storage. Even the reasons you would otherwise buy NetApp or another proprietary storage solution (compression, de-duplication, checksums) is all implemented by ZFS.

    NetApp recently lost their patents based on prior art (they basically ripped off somebody's paper and put in a patent for it), appealed it of course and now they are trying to squeeze the last money out of small shops before they get the smack down from the patent office. This is a very similar case to the Caldera/SCO cases.

    • by kindbud (90044) on Friday July 09, 2010 @05:46PM (#32855648) Homepage

      While people are buying $250,000 NetApp installations, the exact same hardware, performance and connectivity will go for $5000 of high-end hardware and a couple of hours work with ZFS.

      Having evaluated ZFS on a Sun server at that price point, I can state with a good degree of certainty that this is not even close to true. ZFS suffers from several show-stopping performance problems that made it unsuitable as replacement for NetApp filers even on our small applications. Last I checked, the issue we opened with Sun are still outstanding.

      ZFS is probably OK for hosting a file archive, or staging backups. It is far from OK for hosting a Oracle or Mysql database.

      • by guruevi (827432) <evi@smo k i n g c ube.be> on Friday July 09, 2010 @10:02PM (#32857448) Homepage

        There is no magic that NetApp adds that would make it any faster or slower than a comparable system using ZFS. I've seen, implemented and used both systems, they both use commodity hardware. There is no reason your ZFS should work any worse when you get down to the wire than your NetApp appliance unless you're doing something wrong. If you're comparing Sun's RAIDZ with their proprietary RAID0 then you're looking at the wrong implementation for a database. If you add a couple of L2ARC and ZIL SSD caches to your ZFS system, your database should be able to sustain well over 10000 IOPS - on commodity hardware. The other problem with NetApp is it's limits - 16TB is laughable in my field (even 16PB will be laughable in less than a decade).

        When I look at NetApp's white papers, all I see is a similar implementation to ZFS. NVRAM as caches, aggregated writes, snapshots... As I said, they're afraid because ZFS does everything NetApp does and more for a heck of a lot less. NetApp seems to have stopped developing major features which could have distinguished them from other systems and instead just focuses on selling as much systems in the next few years as possible and possibly litigate to get some more cash (the SCO business model).

    • Re: (Score:3, Informative)

      by chef_raekwon (411401)

      While people are buying $250,000 NetApp installations, the exact same hardware, performance and connectivity will go for $5000 of high-end hardware and a couple of hours work with ZFS

      lies. having worked in this industry for far too long, when I see bullshit claims like this, I have to call it. There is no way in hell you will get multiple trays of 15k rpm FCAL storage and redudant FAS controllers for 5k, I dont care if you're Samuel Jackson in the negotiator.

      I run multiple datacenters with emc, hp and ne

  • btrfs successor (Score:5, Interesting)

    by mlts (1038732) * on Friday July 09, 2010 @03:31PM (#32853958)

    Even though btrfs isn't in production yet, we really need a successor to it not just to replace the filesystem, but to replace the LVM layer. ZFS isn't just a filesystem, but also goes one layer lower, coordinating RAID.

    I wish I had the cash to make an open source (GPL or BSD license preferably) bounty for the following in a filesystem/LVM replacement, since ZFS isn't going to be going past Sun hardware these days:

    1: Deduplication on the block level. This would be selectable because in some cases, there would be performance issues to it... but a good filesystem would stick heavily duplicated blocks on fast media (flash or inner cylinders).

    2: 64 bit CRCs. This way, a backup program just has to pull from the filesystem stored CRCs and it would know which files have been changed or not. This also helps with integrity checking.

    3: Compression. Selectable levels would be nice, from a fast zip based to bzip2 -v9.

    4: Encryption, perhaps like EncFS where encrypted directories can be cattached at will. Even better would be more elaborate (public key, smart card) key management.

    5: Block device encryption. It would be nice to install the OS, set a flag that all further writes will be encrypted to a key, then proceed to copy data to the machine. This way, the machine can get set up and (ab)used without waiting for disks to encrypt.

    6: TRIM support. Enough said.

    7: Ability to move data so one directory might be on a three-way mirror, while the rest of the filesystem sits on a RAID-Z equivalent. This way, critical documents are protected.

    8: Advanced snapshotting functionality. It would be great to be able to restore a machine by booting from a USB flash drive or CD, having the filesystem be configured to the hard disks at hand, then copy from a stored image, regardless of architecture or setup of the previous machine's drives were. This way, a machine could be snapshotted, it be moved to a completely different configuration, then restored. A good example of a nice way to restore would be IBM's Sysback utility for AIX, where one can completely redefine where data resides before kicking off a restore.

    9: Advanced attributes, where files can be flagged where if they are unlinked, the OS does a manual TRIM or multiple overwrite, and so on.

    10: Automatic repair of damage. Starting with Windows Server 2008, Windows does a background check to look for damage in mounted NTFS filesystems. This way, something like missing free space or other issues can be flagged before it bites someone in the next bootup. For example, when a machine is idle, it will compare written 64 bit CRCs to what is on disk to ensure that they match, and flag nonmatching files as possibly corrupt.

    11: Ability to add varying amounts of ECC to a filesystem. This way, the volume can take a lot of damage, but the files are highly likely to be still readable. A good example of this is Nero's SecureDisk, where it writes invisible ECC information to burned CDs/DVDs which can be used to piece together damaged files. This way, volumes that are stored for long term archiving can sustain damage, but there is a good chance of recovering the files, or at least knowing the files were damaged.

    • Re: (Score:3, Insightful)

      BTRFS already does multi-disk filesystems including RAID 0 and 1 by itself, bypassing LVM/mdadm AFAIK. RAID 5/6 is a planned feature.

      I don't know about the rest of the stuff you mentioned; I'm not a filesystem guy. Recommend any good books?

    • Re: (Score:3, Insightful)

      by eln (21727)

      but a good filesystem would stick heavily duplicated blocks on fast media (flash or inner cylinders).

      Why? Just because a block of data is duplicated all over your filesystem that doesn't mean it's accessed all that frequently. If I have a disk with 8,000 slightly different ghosted Linux disk images on it, I'm bound to have plenty of blocks that are identical in all of them, so deduplication would save me lots of space. However, since they're probably all just there for archival or testing purposes, and I only occasionally need to access any of them, putting those deduplicated blocks on fast storage woul

    • I wish I had the cash to make an open source (GPL or BSD license preferably) bounty for the following in a filesystem/LVM replacement, since ZFS isn't going to be going past Sun hardware these days:

      And it really hasn't taken off there either. All the Sun shops that I know still use Veritas volume manager and filesystem.

    • Re:btrfs successor (Score:4, Insightful)

      by dgatwood (11270) on Friday July 09, 2010 @04:58PM (#32855136) Journal

      I wish I had the cash to make an open source (GPL or BSD license preferably) bounty for the following in a filesystem/LVM replacement, since ZFS isn't going to be going past Sun hardware these days:

      ...list of requirements...

      To which I would add:

      12. BSD-licensed. What's really needed is a robust filesystem that's ubiquitous---a filesystem that is supported across operating system platforms--- and let's face it, no GPL-licensed filesystem will ever darken the door of any OS except Linux, at least not as part of the OS. Even the *BSDs won't touch a GPLed filesystem except as a from-scratch rewrite. Forget about commercial UNIX vendors. And although you can gain partial support through little tricks like FUSE, such workarounds will never be as fast, as reliable, or as integrated into the whole user experience as a native filesystem would be.

      There's little point in taking any filesystem beyond academic research into a production-quality filesystem if the resulting disks can't be moved from one machine to another without forcing users to being a single-OS shop. Most of the real world doesn't consist of single-OS shops. In fact, an ideal next-gen filesystem would integrate SAN capabilities at the filesystem level so that disks could not just be moved, but actively shared between boxes running Linux, Mac OS X, Solaris, etc. That's just not likely to happen unless you choose a much more open license than the GPL.

      Don't get me wrong, I've written and licensed code under the GPL. There's nothing inherently wrong with the GPL for tools/utilities, GUI apps, and other end-user bits. For libraries, filesystems, drivers, and other code that is tightly coupled with the OS or software built on top of it, however, the GPL is really a rather poor choice. It sends the message, true or not, that you care more about dogma than interoperability, which is generally not a good thing. Ask me why I don't run either btrfs or ZFS.... It's the license.

      • Re:btrfs successor (Score:4, Insightful)

        by Kjella (173770) on Friday July 09, 2010 @09:14PM (#32857232) Homepage

        Meh, personally I'd say LGPL so you don't get slightly incompatible file systems. With a BSD license the temptation is pretty big to keep the changes for yourself and say "works on our OS" while diverging from the common code base. Either that or a very detailed and exhaustive spec with a strict conformance test suite.

    • Re: (Score:3, Interesting)

      by Sinical (14215)

      I'm not gonna go through and do a feature check, but there is Ceph [newdream.net]. It's still pretty early in the development, but looks pretty promising. It uses btrfs as the underlying filesystem.

      Someone put together 1.2PB of Glustre (with dual replication) at my company and it's been problem free so far...

  • This summary makes it sound like ZFS is teetering on the edge of destruction. I thought ZFS was used all over the place by big database warehousing organizations. I went to a PostgreSQL conference a few years back and it seemed like everyone was using it. Is ZFS in such a weak position that one patent troll could have any significant impact on it?

    • No, it's not. In fact Sun is pretty much winning at ever corner against NetApp in the patent fight.

    • Depending on what percentage of their product portfolio Coraid just suspended shipment of, they may be in serious trouble; but that isn't really Sun/Oracle/ZFS's problem...
  • by C_Kode (102755)

    Maybe this is the reason BtrFS is maturing at a breakneck pace. :)

  • by vm (127028) on Friday July 09, 2010 @04:47PM (#32854986)

    It might be worth reading through NetApp's original ZFS patent lawsuit vs Sun [sun.com] before making wild speculations. It seems to me that they are now trying to sue other, smaller companies after their original attempt to sue Sun had failed.

Opportunities are usually disguised as hard work, so most people don't recognize them.

Working...