Submitted by Due to me not working for OSAlert these past eight weeks, I’ve been a bit out of the loop, as I didn’t really follow technology news. I did notice that a lot is going on in OpenSolaris land, and today, Oracle has outlined what it has planned for Solaris 11 – and according to some, the fears about OpenSolaris’ future were justified.
Ever since Oracle acquired Sun last year (it went official early this year), there’s been uncertainty about the future of Solaris, more specifically, about OpenSolaris. Oracle’s John Fowler, Oracle’s executive vice president of systems, spoke with InternetNews.com about the future of Solaris, and it doesn’t look for the open source aspect of it all.
“We have a large new version of Solaris coming out in 2011 called Solaris 11,” Fowler said, “You’ll see us really building on the gold standard capabilities that are in Solaris today with significant enhancements in virtually every aspect of the system.”
What can we expect from Solaris 11? Well, it’s planned for the second half of 2011, and will see major changes all across the system, such as upgrades to “networking stacks, threading, file systems, package management and maintenance”. In addition, a lot of work will be put into scalability to get Solaris 11 ready for a new generation of hardware Oracle is developing.
While all this development is going on, the OpenSolaris community hasn’t been sitting idly by. They are afraid that the lack of communication out of Oracle means OpenSolaris is pretty much a dead end, and as such, they started the Illumos project – while not a fork just yet, it will make it easier to create a fork of OpenSolaris if necessary. Well, what Fowler had to say about OpenSolaris and the open source community didn’t exactly make my unicorn’s horn glow, so that fork might be right around the corner.
“We’re really getting our grip around building a mission critical operating system that includes technologies from openSolaris that we’ve been working on for some time as well as technologies that we’ve more recently developed,” Fowler said, “Solaris 11 will be a superset of what is in openSolaris.”
Simon Phipps, former Chief Open Source Officer at Sun and currently a member of the OpenSolaris Governing Board isn’t particularly hopeful about the future after reading Fowler’s interview. On Twitter he wrote that “Oracle apparently confirms it is no longer developing Solaris as open source”, in reference to this interview.
It would appear to be the case that Oracle is developing stuff for Solaris 11 that will not become available as open source (the “superset” thing), and while that indeed is a bad thing for the OpenSolaris community, I don’t think Oracle’s professional, corporate customers will really care.
Solaris was always a superset of OpenSolaris (all the proprietary bits they threw in).
Still, the lack of any commitment to OpenSolaris is a worrying change from the old Sun propaganda.
Of course. OpenSolaris only came about because Sun wanted to try and tell people Solaris was open source when competing against Linux suppliers, whilst secretly not relinquishing any control whatsoever. So yes, OpenSolaris was always an inferior subset of Solaris.
That strategy obviously hasn’t worked (Sun went bust) so Oracle and Ellison are going in the opposite direction – make Solaris, and SPARC on the hardware side, high end premium platforms to satisfy Ellison’s ego trip of competing against IBM, so they’re the equivalents to AIX and Power. Solaris on x86 is all but dead again once the dust settles.
It’s a somewhat better strategy than the doomed one that Sun had for ten years in all honesty, which was to try and pretend that PCs, cheaper and more powerful x86 hardware and an open source ‘Unix’ in Linux would all disappear when they woke up in the morning.
Of course, customers will care. Nobody wants to be locked into proprietary expensive platforms.
The larger the company, the better informed they are. After SUN defended and presented the virtues of opensource, it will be hard to put the genie back in the bottle, particularly when mission critial support for linux is available and linux remains open source.
Sorry for those folks at SUN that worked hard to move the company forward to see now Oracle move in the complete opposite direction. The thing is that SUN should have focused on hardware and integrating the best features of Solaris into Linux, so as not to shoulder the development of a whole OS on their own, much like IBM and HP do currently.
Many of us never fully gave Opensolaris a shot for fear of what is currently happening. We know it has a sound tecnical basis, but it provides very little that cannot be had in Linux and I like the GPL much better than SUN^A's CDL.
Why should they focus on getting their best features into linux..?
It’s cddl and i personally see it as compromice between the gpl and the bsd. And as a better licens then the gpl. Not that it’s relevant here.
You mean the same customers that are buying Oracle?
Every technology investment comes trade-offs. To make technology decisions based entirely upon whether or not the code is open is foolish. There are plenty of cases where expensive proprietary software pays for itself over time.
Totally agree. Being in the enterprise Solaris game myself I couldn’t care less if it’s open or not. I want it to work and be supported both of which it does this very, very well.
What are you talking about, ” Nobody wants to be locked into proprietary expensive platforms.”
How many corporations are locked into Windows or Apple? And they do not care whether is proprietary or not. Companies just need it to work. It is people like you, that do not understand businesses, that care whether something is GPL, CDL, or whatever. The ones making the important decisions don’t give a hoot about that!
A major corp. will consider Solaris 11 from Oracle a more viable option than Open Solaris from the so called community. Why you may ask? Because corp. are conservative and try to avoid risks at any cost. When your job is on the line, you will choose what is safe and works. Not you though, you make decisions based on passion on a reality that does not exists.
Take a business 101 course, maybe your eyes will open wide, but maybe not!
The only thing that one can take from you post is arrogance.
Bereft of any reasons why a company would bet its livelihood on a on operating system that they cannot fix or control in any meaningful way, the only thing you can add is a personal attack based on your presumed intellectual superiority.
I run a business, a successful one at that. And many of my clients come to me with exactly the same concerns that I have conveyed here. Not small businesses, by the way.
Wonder why Google, Nokia, or HP use Linux internally? Or Hertz or Avis and millions of other companies.
By the way, life is about hedging your bets. You may want to run Oracle, or may have to for some reason, but may prefer doing so on an OS over which you can have some control.
And Linux has greater industry support than Solaris from the likes of IBM, HP, Dell and every major hardware manufacturer or software vendor. The tides have turned, the ship has left. Proprietary Unix has no future.
Dell and HP are now offering Solaris x86:
http://www.goodgearguide.com.au/article/355052/dell_hp_resell_oracl…
Nokia don’t just use linux.
Nokia siemens are one of my employer’s customers, 55K computers running windows, it’s because they still use IE6 that we have to make everything we do work for it.
People were locked into DEC and IBM products because they worked very well and customer support was excellent. They wanted to be locked in. They had the development tools to create their own software and even tho’ the mid to main frames were expensive, they could support many users on VTs or even XTs.
Companies make choices and often they will pay extra money for extra support and/or features that might not be found in open source products.
That is BS. Nobody wants to be locked in, but sometimes companies will reluctantly chose lock in due to lack of better choice.
The thing is, Solaris now doesn’t lot of things better than GNU/Linux, and year from now, it will be behind.
Other thing, Soalris 11 will have huge price tag and only way to get expertise is Oracle university, which is not exactly cheap.
On the other side, there are free distros like Ubuntu, CentOS and Debian GNU/Linux, and every kid can learn that and grow up by using it. Then, getting RHCE is walk in the park.
Solaris is sliding at inevitable death by obscurity.
Agreed your post is BS.
Wrong. Solaris does a lot of stuff better than Linux. Have you heard about ZFS? DTrace? ABI stability? etc etc etc
Why do you say S11 will have a huge price tag? It seems that Solaris 10 is free to run for non commercial use.
And the step to Solaris is quite small, too. There are several Linux companies that switch to Solaris because of limitations and bugs in Linux.
http://blogs.digitar.com/jjww/2008/04/democratizing-storage/
http://lethargy.org/~jesus/writes/choosing-solaris-10-over-linux
And also, the Linux kernel devs say Linux has bad code. Didnt you know? I dont see how bad quality code will catch up technically with Solaris. Maybe market share, yes. But not technically. It will not happen. BTRFS is a prototype. According to blog from august 2010, it is only two full time paid Oracle developers working on BTRFS. Oracle seems to not be serious with BTRFS? Where is the large dedicated BTRFS team with lots of resources? There are none? There is instead an ZFS team, that sells machines today? Ok. Then maybe Oracle should kill off BTRFS?
Linux Kernel dev David Miller:
http://kerneltrap.org/Linux/Active_Merge_Windows
“The [linux source code] tree breaks every day, and it’s becomming an extremely non-fun environment to work in.
We need to slow down the merging, we need to review things more, we need people to test their f–king changes!”
Andrew Morton:
http://lwn.net/Articles/285088/
Q: Is it your opinion that the quality of the kernel is in decline? Most developers seem to be pretty sanguine about the overall quality problem…
A: I used to think it was in decline, and I think that I might think that it still is. I see so many regressions which we never fix.
And Dave Jones
http://www.kroah.com/log/linux/ols_2006_keynote.html
“Last year Dave Jones told everyone that the kernel was going to pieces, with loads of bugs being found and no end in sight.”
Maybe you have missed the discussion where Alan Cox quits as a developer because Alan argues that the Linux regressions should be fixed correctly, which may break user applications? And Linus says that if user applications breaks, then you should not fix that Kernel issue correctly. Instead you should preserve the old behavior so user apps doesnt break. Alan complains on the Linux bugs, Linus says he shouldnt mind them.
http://lkml.org/lkml/2009/7/24/182
http://lkml.org/lkml/2009/7/28/375
“Quite frankly, I don’t understand why I should even have to bring these issues up. You should have tried to fix the problem immediately, without arguing against fixing the kernel. Without blaming user space. Without making idiotic excuses for bad kernel behavior.
The fact is, breaking regular user applications is simply not acceptable. Trying to blame kernel breakage on the app being “buggy” is not ok. And arguing for almost a week against fixing it – that’s just crazy.
Linus”
And Linus T says something like “Linux is bloated”, “The I/O foot print is scary”. etc.
And of course, it is much easier to find non-linux users that talk about the bad code of Linux. For instance FreeBSD programmer Theo de Raadt
http://www.forbes.com/2005/06/16/linux-bsd-unix-cz_dl_0616theo.html
“[Linux is] terrible,” De Raadt says. “Everyone is using it, and they don’t realize how bad it is. And the Linux people will just stick with it and add to it rather than stepping back and saying, ‘This is garbage and we should fix it.'”
Another CEO for a software company:
“You know what I found? Right in the [Linux] kernel, in the heart of the operating system, I found a developer’s comment that said, ‘Does this belong here?’ “Lok says. “What kind of confidence does that inspire? Right then I knew it was time to switch.”
So… “Solaris now doesn’t lot of things better than GNU/Linux, and year from now, it will be behind”? You really believe it, yes? :o)
You are wrong about everything, but I don’t know if I have time to refute everything. thare is a lot of nonsense in your post.
Solaris 10 if free to use for 90 days. Affter that, pay to Larry or loose it. You obviously missed the news.
Again, yum missed the news. OpenSolaris is not a viable choice anymore. It is now updated for more than a year and it will never be updated. And only ILL LOONS would use ILLUNOS in production. It will probably be incompatible with Solaris. For Solaris, you have to pay up license and maintenance contract. Or reinstall every 90 days.
Wrt Theo Theo de Raadt: He is not, and never was, a FreeBSD developer. He is resident troll of OpenBSD mailing list and before that, he worked on NetBSD until they kicked him out because of his inappropriate behavior and trollish nature. Here is the announcement
http://mail-index.netbsd.org/netbsd-users/1994/12/23/0000.html
If you listen to him, I fear you are no better than him. He is just buthurted because nobody cares about OpenBSD.
Wrt kernel developers: That is called OPEN AND COLLABORATIVE DEVELOPMENT. Something that Sun never had, so you don’t know how it looks. Sun was always hushed their devs from talking about nasty flaws in Solaris. But in Free Wold, it is natural that people argue about quality standards. Actual kernels that go into mission critical distros like RHEL are extensively hardened before released. Kernel.org Linux 2.6.xx was never release quality software, it was always beta. Release quality comes with 2.6.xx.yy minor releases (notice yy) and after that, distributor test it and patch it even more and it goes like 2.6.32-35el (el for enterprise linux by Red Hat and 35 for number of updare). So lkml posts don’t speak about quality about “production Linux”; it speaks about quality of development branch. Which was never high, but neither was OpenSolaris-dev.
As for “BTRFS is prototype” Well,then is ZFS too. ZFS is developed by Jeff Bonwick and two other guys working full time on it. Later continued with little testing and input from couple of other Sun guys.
BTRFS has 3 full time engineers, two at Oracle and one at Red Hat. Of course, with help of rest of Linux community. Also note that ZFS is “rampant layering violation”, meaning that puts everything and kitchen sink on top of VFS layer which is not elegant solution and require lots of work. It has lots cut-corners and kludges.
BTRFS is file system only. It is designed to be a file system and there is LVM for RAID and other stuff. That means that BTRFS is smaller and cleaner and needs less man-hours to get to production quality. And it gets all the features of ZFS (and then some!), thanks to smart layering and using user space utilities. Oracle can’t kill BTRFS, Red Hat will simply hire the devs and do it themselves.
Did I miss anything? I think I just proved you wrong about everything.
Yes man, you really proved me wrong on everything. I remember your earlier attacks on me, and you really “proved me wrong” back then too.
http://opensolaris.org/jive/thread.jspa?threadID=132608&tstart=15
The post by Edward Ned Harvey, Aug 1, 2010 3:15 PM
He says Solaris 10 is free for evaluation use, research and instructional use, etc.
And also, you can buy Solaris 10 for 20 USD.
https://shop.oracle.com/pls/ostore/product?p1=OracleSolaris10MediaPa…
So where is this “huge price tag” you talk about? It seems it is you that missed the news?
I did not talk about OpenSolaris. I talked about Solaris 10.
You know, Solaris is not like Linux, where you have different incompatible package managers. You are just guessing or wishing here. Or FUDing.
As I showed you, no. You are free to run Solaris 10. But if you want to have support contract, then you pay.
He is a developer, amongst many, that says the Linux code sucks big time. I gave other links including Linux kernel developers. Should we listen to them when they say Linux code is bad, but not to Theo?
Maybe you missed the Solaris problems posts, in the OpenSolaris forums? Why has not them been censored?
The code is open, I have not heard about ANY developers complaining on bad code in Solaris. But there are lots of developers including Linux devs, that complain about the bad Linux quality.
I also gave other links than to lkml. For instance, The God Linus T explaining that Linux is bloated. And Intel studies showed that performance dropped for each new release. Something like 10% or so.
Cool. I didnt knew this. Where is the link? Or is it just guesses or FUD again?
ZFS is not layering violation. It has only less layers, as chief architect explains. I also heard that BTRFS is violating the layers. I dont know if it is true, but I read it.
Regarding ZFS has lots of “kludges” etc, I didnt know that. Are you making this up, or do you have link? In other forums, there are lots of Linux fanboys that FUD and make up things and lie. And they _confess_ they FUD, and they _confess_ they lie, and I prove sometimes that they lie by supplying a link. What is the problem with you Linux guys? Can’t you speak true things? I link to credible sources, such as Linux kernel developers. I do not make up things, nor Lie.
It sounds as you guess again. ZFS is very neat. It was 75k LoC some time ago. Which is extremely small. This can be achieved because ZFS has scrapped several unnecessary layers, that other filesystems have for legacy reasons.
What do you mean, smart layering? Is BTRFS also violating the layers, just as I read it does?
I am not saying that Oracle will kill BTRFS. I am only saying, from a business perspective, it is better to make sure BTRFS is lesser copy of ZFS and never surpass ZFS. If BTRFS is better than ZFS, then Oracle can not sell all those fancy ZFS storage servers.
Correction: I meant of course that performance dropped 10% over the last releases. Not every release.
I do not want to state false things, would I?
http://www.theregister.co.uk/2009/09/22/linus_torvalds_linux_bloate…
“Citing an internal Intel study that tracked kernel releases, Bottomley said Linux performance had dropped about two per centage points at every release, for a cumulative drop of about 12 per cent over the last ten releases. “Is this a problem?” he asked.”
ZFS is not a rampant layering violation, it is a re-organisation of the layers.
Instead of 8 or 9 or more layers between the apps and the disk, there are now 4 (physical, DMU, ZPU, filesystem, or something along those lines). Aka, stepping back, looking at the current state of storage, realising that the 30-year-old assumptions are no longer correct, and coming up with something better.
Btrfs may have started as a simple filesystem only. But it’s now a filesystem, a volume manager, and a RAID engine. With more on the way.
Nope, the volume management and RAID is built into Btrfs now. Just like ZFS.
If only that were true. I’m guessing you don’t read the btrfs-devel mailing list. Or you’d realise just how funny your comment is.
Wow, you are so far out in left field on this, that it’s no longer funny. Btrfs is still missing half of the features of ZFS:
single-parity raid (aka raidz1)
dual-parity raid (aka raidz2)
triple-parity raid (aka raidz3)
deduplication
encryption
multiple compression schemes
dataset streaming (zfs send/recv)
There’s more, but that’s a good start on the ZFS features that Btrfs is currently missing (with no plans for incorporation in the next 6-12 months).
Actually, other than some corrections about Theo, you’ve proven your lack of factual information quite well.
Maybe I worded my post poorly. Basically that is what I am saying. Companies CHOSE to be locked into specific hardware/software combinations because those products met their needs and did a good job of it… and STILL DO in many cases. VMS is still supported by HP because there is a sizable install base and people don’t WANT to switch.
He gave a reason:
Companies just need it to work.
Companies also buy photo copying machines and air conditioners that they cannot fix themselves. The vast majority of companies that run Linux don’t work on the OS either. They pay a company like Red Hat and focus on their own business.
And what if there is a technical benefit when it is tied to Solaris on Sparc stacks? Just ignore any productivity gain and focus on a feature that most companies don’t care about?
Oracle isn’t really in the OS (or even hardware) business, they are in the turnkey solutions business. If you’re going to run apache, postfix and Postgres on your commodity hardware servers then they have no interest in you as an OS customer.
The customers they do want are the ones who will run Oracle software on the Oracle(Sun) operating system on Oracle(Sun) hardware, all wrapped up in a juicy Oracle service contract. They want to sell you the whole stack, and if you aren’t interested in the whole stack then they’re less interested in you as a customer.
I like your well thought out post. Oracle can’t do much about the low cost or free solutions. Those will always exist and Oracle may not want to stick their necks out there to try and offer free solutions for which it can’t monetize. They do quite fine with enterprise solutions and the support from them.
The benefits of Oracle over free databases are the features, support and experience Oracle has in the industry. They can debug from app server to DB what you need and offer solutions including patches.
What they have to offer on the hardware and OS side is exactly the same:
The benefit of Solaris over free OS are the features, support and and experience Oracle has over the industry. They can debug from hardware to kernel what you need and offer solutions including patches.
That kind of support is one that they don’t have to guess at. And I’ve received many OS related patches from Sun in the past for specific problems that were not even announced until they became GA.
Let’s also not forget transactions per second. Oracle will just plain outperform MySql.
How many companies actually need Oracle is a valid question but enough of them don’t mind paying for what is perceived to be the best when it comes to performance and reliability. For a big Fortune 500 corp a 100k check to Oracle is pocket change.
I wonder how do you imagine that people are doing business today. Do you think that companies edit and recompile the source code of the servers they run their business on ? Open Source or not makes little or no difference.
If some company decide they want to use open source product, they are likely to go to vendor like Novell or Red Hat. If they do, they will very pay steep price, and very much depend on support from those companies.
There is a difference if company is using modified OS as a part of their business process, like some who are selling devices running embedded Linux. But, such companies are small percentage.
I have been doing business with the company which is major Red Hat partner in the region. There is no much difference between the way they do business, and, for example some Microsoft partners.
I am 100% sure that lack of Oracle involvement in Open Solaris would not lose them any significant customer.
Why should they pay their developers good money, only to have some guy being able to download OS for free ? It is a waste of resources. Unlike Linux, major part of Solaris have been developed in-house. Contributions have limited significance.
http://news.cnet.com/2008-1082_3-5065859.html
They bought Red Hat once, and hired good sysadmins to take over once they adjusted.
They’re not alone, I know a few sysadmins that are in a similar situation.
Granted, one of them works at a company where they still use dumb terminals attached to AIX machines, but they also have workstations as well, and these run free alternatives to their former Red Hat/SuSE.
That’s about it.
There are pointy-haired bosses, and there are savvy ones.
Both hire people to maintain systems, some just do it smarter. (oh, snap)
Because that guy may want to develop a proof-of-concept to convince his boss the idea works.
There are many reasons, and most of them have to do with getting people to put their hands on the product. Linux has better mindshare right now then Solaris does, and that’s not going to change as long as Oracle locks Solaris up in an gilded tower.
Read all about it! Company that charges obscene amounts for closed software wants to write closed software and charge for it.
From a user perspective, they are stuck in a rut.
Now this is not all bad news.
The run means they are focussed on working with their H/W platform very efficiently.
The bad side (for Solaris) is the X86-64 port. The lack of hardware support means it just can’t stack up against Linux.
If the likes of RH had not canned the Sparc version of their mainstream server product I’m sure that more Linux shops would be using Sparc H/W.
What I am dreading is the price hike that even our Sun, sorry Oracle Salesman says is coming. They’ve already hiked the price of Oracle on our Sparc boxes.
If they do this then even IBM X series blades start to look attractive. At the moment, we run 8cpu, 8core/cpu systems. We could do with a bit more clock speed though but otherwise, they perform pretty solidly.
Do not forget that Oracle do invest in Linux development, so maybe its not so bad that they reduce duplicate efforts and rather focus on Solaris as a enterprise OS for special requirements and focus on Linux for their opensource efforts. Remember that Oracle can only win by Linux gaining market shares… its far too easy for companies to just use MS SQL Server instead of Oracle DBMS with their Windows-platform (really; many of my clients have switched from Oracle DBMS to SQL Server the last couple of years because they use Microsoft for everything else). OpenSolaris on the other hand probably costs more than Oracle can gain from it.
Just think about it; what is really OpenSolaris’s advantage that are not covered by either Solaris or Linux? Will enough people care about those? Truth be told, A LOT of people use Solaris, and a lot of people use linux… I have yet to see anyone use OpenSolaris for anything more than playing around.
As long as they keep investing in Linux and in Solaris, I don’t care about OpenSolaris at all. Its better that they do this rather then end up like Sun, investing a lot of money in technology that is going nowhere…no matter how cool that makes them or how theoretically excellent the technology might be.
Edited 2010-08-11 21:28 UTC
I would say the advantage not covered is the combination of Linux and Solaris, is that OpenSolaris had a free stable production quality ZFS implementation.
Linux really doen’t have anything that matches OpenSolaris ZFS in ease of use and features.
Nothing that is production ready. Wait a year and BTRFS will be here. If you are looking to the future with your data storage needs, BTRFS is a really good reason to *not* go to ZFS, unless you are already running Solaris everywhere.
Who makes BTRFS? Oracle.
While the project was started by Oracle, the code is GPL and already merged int the mainline kernel. Josef Bacik of Red hat is one of the fulltime developers ( along with two others from oracle ). There is widespread plans for adoption from many distros. While an outright Oracle drop of the project would delay it, I doubt they would be able to stop it.
I take it the implication there is that Oracle might stop development? The code is already in mainline and is also being developed by Red Hat and others so Oracle aren’t going to put the genie back in that bottle.
Btrfs is at least 5 years off from being a replacement for ZFS as it stands today. Btrfs is not anywhere near the same league as ZFS right now. It’s absolutely hilarious when people try to suggest “some future version of” Btrfs as a viable alternative for ZFS of today.
Technology always moves faster than you think. In planning adoption for a storage technology, you’d have to be completely naive to not evaluate the future of the product you are going to use for your valuable data. There is a good reason why people dropped plans for supporting Reiser 4. If you really look at the design of the two File Systems, its hard not to be bullish on BTRFS. Everyone who is already on Linux will also have ZFS like capabilities on all their storage devices. That’s huge.
As for timing I guessed a year. Maybe two, depending on your intended use. If Ubuntu makes it their default file system, as suggested, You’ll get a lot of bug fixes pretty quickly. Just hope they submit those back to the mainline
Until it gets support for RAID levels above 1, it’ll be hard to even consider it for a lot of storage applications. The patches for this have been “just around the corner” for over a year now.
Until it gets support for deduplication, it’ll be hard to consider it for a lot of storage applications. There aren’t even any plans for adding this at this time.
Until it gets support for proper volume management (or better integration with LVM/md) it won’t be a good fit for use in storage systems with 10/20/30/hundreds of disks.
Actually, until Linux gets a storage stack comparable to ZFS in OSol/FreeBSD or GEOM in FreeBSD, it’s not a good fit for many storage applications.
Maybe for desktops with 1-4 drives, or for laptops with 1-2 drives, or for small servers with 1-4 drives, Btrfs may be useful (nothing but RAID10). But for large storage servers with 24+ drive bays, or for storage arrays that can handle multiple 48-bay enclosures, it’s just nowhere near ready.
Anyone who believes otherwise is seriously deluding themselves.
Ahh.. Now we understand each other. I think its massive adoption on “small” servers will drive its adoption and over take ZFS on larger systems. That’s the way it works. Get something used by the masses, and it finds its way everywhere into everything. Linux, windows, x86, English, Rap Music
Linux is already in all of those locations sunshine, and anything that large will be using hardware RAID of some kind so ZFS as anything more than a basic filesystem doesn’t even come into it. When it’s ready Btrfs will merely start taking over the role that other storage layers are performing in Linux right now.
The laughable thing is you’re making it sound as if Solaris and ZFS is going to have to be ‘replaced’. Solaris already has been and ZFS has a rather pitiful installed based when it comes to storage.
The other issue to consider is that there is a large industry around ZFS-based storage solutions, and absolutely 0 companies using Btrfs as the basis for a storage solution. Someone looking for a turnkey solution, for a simple “plug it in and go” box, it’s easy to end up with a box that uses ZFS internally (Nexenta, GreenBytes, even FreeNAS, for example). There’s nothing like that available for Btrfs.
Is there? I must have blinked because I missed it.
The storage industry consists of established players using things like VxFX and clustered filesystems. Sun came late into the storage market with ZFS hoping to capture some of that established market share as a cheaper alternative, and they’ve had rather mixed success.
Hmmmm, so that’s what you consider to be a ‘storage industry’? I’m afraid installing Nexenta or FreeNAS yourself is not a storage industry.
All (and I mean all) of the ‘turnkey’, small and cheap commercial NAS and storage boxes I have seen use Linux and generally XFS as filesystem. Not a single one runs Solaris or ZFS. Btrfs already has a ready and established market that Linux itself is already in.
Who’s talking about “small and cheap” or NAS? That’s the home market, not the commercial market.
If it has fewer than 12 drives in it, it’s not a storage box worth mentioning here.
Btrfs is nowhere to be found when you get into real storage solutions. The NetApp, EMC, Veritas, Greenbytes market, not the home “2-drives in a box with an ethernet port” market.
You can spend a couple hundred grand for an iSCSI/FC box from NetApp or EMC. Or less for a Greenbytes box using ZFS+. Or less for a Solaris box using ZFS. Or even less for a custom white-box with OSol/FreeBSD using ZFS.
I’ve yet to see a single product in this area that boasts about using Linux and/or Btrfs.
Sure, maybe Linux has taken over the “2 drives in a box with an ethernet port” market. But who cares? That’s not was enterprises, businesses (even medium-sized ones), or even school districts are looking at.
You specifically used the phrase ‘turnkey’ and started talking about FreeNAS and Nexenta probably because you know Solaris and ZFS aren’t there at that end of the market, and it’s more of a small business storage market these days.
Neither is ZFS. NetApp, EMC and Veritas are the only companies worth mentioning there and Sun with ZFS are merely bit-part players trying to get a piece of the action with something cheaper. They are irrelevant.
You’re trying to create some ‘enterprise storage market here that Solaris and ZFS are a part of and it just isn’t there.
That’s the small and cheap NAS market that you claim that we’re not talking about where Solaris and ZFS are not present. It’s all dominated by Linux systems.
You wouldn’t. They use Linux and XFS, LVM and RAID generally. That will change with Btrfs. Linux is already there. It’s pointless talking about Btrfs because it will appear by default.
Well, yes, that’s what they are looking at because one of Sun’s goals for ZFS was to make dynamic storage cheaper as storage hardware has got cheaper. Unfortunately for Sun they’re caught somewhere in between NetApp, Veritas and EMC and all those smaller 4, 8 12 or more disk storage boxes from manufacturers that all run Linux, just like they were with PC hardware. That’s probably why Sun hasn’t done too well.
Edited 2010-08-13 18:08 UTC
Apparently, NetApp is dead scared of open ZFS, as NetApp sued Sun and tried to stop ZFS. NetApp has some expensive storage servers, and if ZFS can do everything for free or a much lower price, I understand why NetApp sued Sun. Just recently NetApp sued another ZFS based storage company, according to the news. Maybe you missed that. There are several storage companies selling ZFS based machines. For instance, OnStor.
http://www.channelinsider.com/c/a/Storage/OnStor-Pantera-Storage-LS…
I think it would be better if NetApp spent all resources on coming up with new cool tech, instead of spending them by sueing others. NetApp sued Sun in a state famously used by Patent Trolls, despite the fact that Sun and NetApp headquarters are close to each other in another state.
Personally, I can not recommend XFS and Raid solutions, as they are unsafe. Your data is not safe with them solutions.
Here is a link to a PhD thesis, where the researcher concludes that XFS, JFS, ReiserFS, etc are really bad with respect to data safety:
http://www.zdnet.com/blog/storage/how-microsoft-puts-your-data-at-r…
“All the file systems shared…ad hoc failure handling and a great deal of illogical inconsistency in failure policy…such inconsistency leads to substantially different detection and recovery strategies under similar fault scenarios, resulting in unpredictable and often undesirable fault-handling strategies…We observe little tolerance to transient failures;…none of the file systems can recover from partial disk failures, due to a lack of in-disk redundancy. ”
Avoid XFS and JFS.
And Raid5 and Raid6 are not better. There is a whole site with lots of technical articles by sysadmins, dedicated to explain how bad Raid5 is.
http://www.baarf.com/
More research on Raid:
http://www.cs.wisc.edu/adsl/Publications/corruption-fast08.pdf
“Detecting and recovering from data corruption requires protection techniques beyond those provided by the disk drive. In fact, basic protection schemes such as RAID [13] may also be unable to detect these problems.
..
As we discuss later, checksums do not protect against all forms of corruption”
Research shows that Raid6 is also bad:
http://kernel.org/pub/linux/kernel/people/hpa/raid6.pdf
“The paper explains that the best RAID-6 can do is use probabilistic methods to distinguish between single and dual-disk corruption, eg. “there are 95% chances it is single-disk corruption so I am going to fix it assuming that, but there are 5% chances I am going to actually corrupt more data, I just can’t tell”. I wouldn’t want to rely on a RAID controller that takes gambles :-)”
I have much more research articles on this. Just tell me if you want to see read more on how unsafe your data is.
From above examples with all common filesystems ext3, ReiserFS, NTFS, and hardware raid5, etc – you can not just add some checksums all over the place into BTRFS and expect to get data integrity. It is very hard to do it correctly.
But ZFS is another thing! Researchers tried out ZFS and it passed all tests. This is the single reason to use ZFS: your data is much safer. The performance is not interesting. Is your data safe? Here is a link to a research paper on the safety of ZFS:
http://www.zdnet.com/blog/storage/zfs-data-integrity-tested/811
I know. Btrfs might actually have a fighting chance of running on a NAS box.
You do know btrfs is *only* a filesystem right?
ZFS is a rampant layering violation
ZFS is more than just an fs, its a volume manager too. How are you going to RAID your btrfs? Through mdadm? Theres a good reason not to use Linux in the enterprise straight away.
Through bultin multidevice support (raid0,1,10,5,6) in btrfs. I suggest you should educate yourself a little before posting rants: http://events.linuxfoundation.org/slides/2010/linuxcon2010_mason.pd…
No, ZFS is a re-orginisation of the layers to better fit the use-cases of today, instead of carrying on with the out-dated layering schemes from 30 years ago.
Big difference.
There are still layers in ZFS. But there’s only 4 instead of the 8 or 9 that “traditional” storage stacks have.
Think of it as the TCP/IP of the storage world. Afterall, TCP/IP is very much a “rampant layering violation” compared to the OSI network model. Yet no one has any issues with using it, and many even find it to be a better stack than the OSI one.
Btrfs supports RAID0, RAID1, and RAID10 directly “in the filesystem”. It’s even part of the mkfs.btrfs command (now, which is the real layering violation?). And Btrfs supports sub-volumes, making it a volume manager.
The problem is that it’s not all that well integrated with the rest of the system (md/lvm, etc). And it doesn’t support RAID levels above 1. And the tools for working with it are only just now getting to be usable (finally, a single btrfsctl command for everything, similar to the zfs command).
Give it 5 years, and it might be a usable alternative. Hopefully, by then, the whole Linux storage stack will be usable in enterprise-y situations.
btrfs is limited by the laws of physics. it cant catch up to zfs in a year.
assume the purchase of sun by oracle has delayed zfs development by a year. btrfs still has multiple years of development to go before it matches zfs features and stability.
Oracle did invest in Linux development (and had their own distro), before they had their own in house OS to run their software on. Now that they have their own OS and no longer really need Linux I’d be very surprised if the investments continue. I don’t think Oracle cares too much about customers running smallish databases on cheap hardware.
Would a different license to the one applied to opensolaris by Sun have protected it from Oracle? Ie improving it but taking it away from the community.
I remember criticism of Sun for using a cddl license – would someothing like GPL have protected opensolaris from Oracle?
It isn’t the license that is the problem. The CDDL allows for forking.
The problem has long been that OpenSolaris has not attracted outside development. The vast majority of the commits have been from Sun.
Any fork of OpenSolaris would just stagnate.
You needed a sponsor from SUN and the hole progress of getting a patch committed and approved to OpenSolaris was daunting.
This is also what Illumos is trying to tackle. Keeping the standard of code high but actually letting people develop for it without the very reasonable fair of all there work being for nothing.
Nobody wanted to develop for OpenSolaris if the changes of your hard work actually getting excepted are slim to none.
No the problem has been a general lack of interest as can be seen by the relatively sparse OpenSolaris forums.
There is not only a lack of developers but also users.
OpenSolaris has technical advantages but no one cares. Linux is “good enough tech” and there is also FreeBSD.
This is coming from someone that wanted to see OpenSolaris do well. Sun was too late to bring Solaris to X86. You can thank McNealy and his arrogance for that.
“Linux is a “great environment for the hobbyist” but not for corporate IT shops.” – McNealy in 2003
Did the GPL “protect” MySQL from Sun/Oracle?
Well, there’s a subtle difference here: MySQL is GPL + copyright assignment. If it were GPL-only, no proprietary forks/branches were possible. On the other hand, it would become impossible to use MySQL in software that does not have a GPL-compatible license as well.
Edited 2010-08-12 06:53 UTC
I suspect most criticism is from open source advocates that bitch and moan when a company dare not respect their religion.
The former Sun CEO embraced the healing powers of the source and look at how far that got him.
Oracle has a highly profitable business model built around a single proprietary product so you can’t get upset if they don’t want to follow in the same path.
FreeBSD has both dtrace and zfs so who really cares if OpenSolaris goes to the OS graveyard.
I like FreeBSD, but the Solaris kernel scales better.
Edited 2010-08-11 22:00 UTC
I’m still not seeing a good case for OpenSolaris.
Anyone who needs a level of scaling greater than what FreeBSD can offer can afford a Solaris service contract.
Open development is democracy. Democracy is politics, not religion.
There is nothing wrong with defending democracy.
what the shit. open development is not necessarily democracy
Edited 2010-08-12 17:13 UTC
It’s never a religion and that’s my point.
Your insult does nothing to counter my point that religion is not involved.
That must be why Linus describes himself as a benevolent dictator.
Or why Shuttleworth asks the community for their opinion even though he clearly doesn’t care if the majority is against him.
Indeed. Ubuntu is an open source project, but not developed in an open way.
Most upstream projects Ubuntu uses have an open development model (eg GNOME) but Ubuntu itself has not.
Open source is a licensing scheme.
Open development is an interacting method between developers and potential developers.
Neither is a religion.
Or just SOS for short.
That works on a few levels.
^A· ^A· ^A· ^aEUR” ^aEUR” ^aEUR” ^A· ^A· ^A·
(10 points to the first person that figures that out)
i see what yo did ther , +1 internet to you sir
…. — .– .- .-. . -.– — ..- –. — .. -. –. – — …- — – . — . .—- —– – .. — . ..- .–.
HOWAREYOUGOINGTOVOTEME10TIMEUP
Funny!
SOS (in morse code!)
It shouldn’t be hard considering the answer is in the parent.
I wonder therefore where Illumos sits in this when one considers whether Oracle will provide patches back to the community or will Solaris and Illumos be developed independently of each other? As much as I admire the work that the Illumos developers are doing I am concerned that Illumos will always be several steps behind Solaris in terms of not only features but code patches as well.
Personally I find the whole attitude regarding the shunting of OpenSolaris to the side seems pretty defeatist if you ask me – they still have failed to address the underlying problem with Solaris and simply recoiling back into proprietary was of doing things isn’t going to solve the problem. Solaris’s lack of hardware support for example, ancient subsystems finally being replaced (boomer, wireless stack etc) and support for new hardware. The attitude of ‘only focus on the server’ is also doomed to fail – but retreating further and further up the enterprise ladder than addressing problems seems to be a tradition that’ll continue on long after the Sun brand becomes a distant memory.
Why are people critical of Oracle’s move here. OpenSolaris has been around for a while and it garnered little interest. If most of the OpenSolaris contributors were already Sun employees then what is the point. There were those who wanted Solaris to be open sourced just so that they can add the good parts to Linux. Why would a company invest billions on a product to just give it away. How many people work for free.
I never liked the move to make Solaris open sourced in the first place. No one is banging on IBM’s door (or HP) to make their UNIX open sourced so why did Sun have to get a bad reputation. Is it because Solaris was just a damm good OS and was popular? I love Solaris (and I also love Linux, and *BSDs), it was a clean OS with lots of developer tools and a great JumpStart system for enterprise deployment.
People are willing to pay good money for a 24×7 supported product.
Or rather Sun were hoping others would work for free and contribute to Solaris like they do for Linux. When it didn’t happen, Oracle thought screw you and went back to closed and proprietary software model. Well, there probably were other reasons too.
It just goes to show that “open source” is a buzz phrase, and if open source does not make money, it will be tossed aside for something else. Whenever a corporation “embraces” open source, they are just looking how to make more money quickly and get someone else to write code for free.
The system resembles a rich landowner that has peasants working on his fields. The peasants are given open access to the fields, however the landowner is the one who profits the most from their work.
From the improvement list it sound like a Red Hat Linux like development focus. Enterprise, Enterprise.
Would love them also focussing on the Home user market.
But who can blame them seem’s it’s the only real revenue driver for Linux and Solaris.
forking opensolaris would be a personal exercise by the forkers. the project is too large to actually survive with so little blood. eventually they’d get bored and quit.
all potential opensolaris fork developers should save some time and help the greater good by joining an existing unix-like project instead. I hear freebsd and pc-bsd have work that needs to be done.
FreeBSD 4 was forked as DragonFly BSD with fewer developers.
such a modest comparison must thrill potential forkers
… and since then they developed quite few interesting bits:
http://www.dragonflybsd.org/features/
From the article:
“A focus on scalability would also be incorporated in preparation for the next generation of hardware, like the 128 core, 16,384 thread system with 64TB of memory that Oracle is currently developing.”
This means that Oracle os going neck in neck with IBM zEnterpise and HP Integrity Superdome2 mainframe platforms. Thay want to Solaris to become big player in a big game.
And the question why can be answered from this article:
http://arstechnica.com/hardware/news/2007/07/legacy-matters-why-the…
“profit margins on a product [System Z mainframe] that are rumored to be as high as 85 percent”
So Oracle has a wet dream about similar profit numbers where they can lock the customer into their network of contracts, hardware, software, spare parts and support.
BTW no wander, IBM has such a profits: the System Z contains a lots of Power CPU-s, but they are devided between different groups, and Java code for example can be run only with specifically taged CPU-s. When customer wants to increase their Java performance on the mainframe, they must purchase specifically taged (althoug regular otherwise) POWER CPU from IBM. Even if they have available CPU resource in the mainframe they cannot run Java on those CPU-s because those are not marked as appropriate.
Thats called vendor lockin and corporate world is pretty happy about it becase the work gets done. The customers are not able to jump the boat usually because the jump would not make sense financially. They have invested too much money into the infrastructure to replace it all.
And that is the world, that Oracle wants a piece. Thay want money from rich enterprise customers. And Solaris + SPARC platform is the key into thar wanderland.
It is not POWER. Mainframes don’t use POWER, they use s390x a.k.a z/Architecture. And specialty engines are not required to run Java or Linux. They run just fine on general purpose mainframe engine. Thing is, specialty engines are cheaper, but they have certain instructions disabled. So you can’t run z/OS on Linux specialty engine, or on Java specialty engine. But they run Java and Linux same as GP engine and have bigger bang for a buck since they are cheaper. For z/OS you have to buy most expensive engines. Here is more about mainframes.
http://en.wikipedia.org/wiki/IBM_System_z10
New mainframes which are coming out this year will use POWER CPUs too, but only as accelerators. Main CPUs will only use s390x.
And about lockin… I know CIOs who got into mainframe game. Believe me, they are not happy. They curse the day they thought about Mainframe if they are ones who made the decision, and those who are inherited that state the union are cursing whichever CIO come up with idea to sell company to IBMslavery.
Good CIOs don’t lock-in. Ones who experienced it, hate it. They will accept it sometimes when they perceive there is no better choice, or when they fail to see lock-in trap in plain sight (this is how Microsoft gets its customers: through deception. They say: it is widely used “industry standard” and some people think that means “no-lockin”). Most regret when they see that they put their head in the noose. Then they make efforts to reinvent their whole IT infrastructure and move to open source. But it is hard. It is not called lock-in for nothing.
Free Software looks like some strange ideology at fist, but when you get burned repeatedly by money grabbing proprietary companies, Stallman’s ideology starts to make very practical sense and all the sudden starts looking very pragmatic.
From the IBM website:
“The zEnterprise 196 each MCM features six 5.2 GHz Quad Core processor chips” (up to 4 processor books, each containing an MCM)
I assumed they were POWER7 chips
Of course System Z mainframe has also general purpose CPU which coordinates the task distribution.
Anyway if IBM uses mix of different chips in their mainframes, HP atleast is dedicated to Intel Itanium chips that not many find a use for
Actually HP Superdone2 seems just a rack of small computers, each containing CPU and memory.
It doesnt matter which CPUs IBM Mainframes use. The Mainframes lag behind cpu wise. They are dog slooooow. As I have shown earlier with links and calculations, you need a PC with 8-way Intel Nehalem-EX cpus, to match the biggest IBM Mainframe with 64 cpus. That is horribly bad performance from the Mainframe. In fact, you can emulate a Mainframe on your laptop with “Herkules”.
But the biggest IBM Mainframe with 64 cpus, probably costs a couple of 10s of million USD.
If someone asks me, I can show these links and calculations again. Please, someone ask me to start over again about how slow the Mainframe cpus are. :o)