Sun Microsystems has delayed the release of two major features the company has trumpeted as reasons to try its latest version of the Solaris operating system. Eric Schrock, a Solaris kernel programmer, said on his blog in April that he’s “completely redesigning the ZFS commands from the ground up” after finding some deficiencies.
Sun delays major Solaris features indefinitely over deficiencies
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Hi,
I’m annoyed as the next person, and I’m kinda annoyed at Sun…but if you read the article carefuly, they do say that it will be available via Solaris Express sometime this year.
I’m hoping it’ll be in the next few months, as these two features are some of the most attractive features of Sol10 for me – ZFS mainly for the coolness factor (somehow I doubt I’ll come into a 10Tb disk pool anytime soon *grin*), and Janus, for actually being useful.
*fingers crossed* (if any Sun people read this, could they provide a rough idea of which SolExp build it’ll be?)
bye,
Victor
Waiting for “Show me the Solaris or dont talk about it guy”!!
ZFS doesn’t ship, WinFS doesn’t ship.
Funny old world. Of course, ZFS seems real, whereas WinFS seems to be second system writ large
Doesn’t really go into much detail, then again, I am not surprised, it isn’t as though it is a minor piece of technology; it is a critical piece of technology that must be done correctly or otherwise serious problems could occur.
But with that being said, firing programmers each quarter isn’t exactly going to make the job any easier.
It will be interesting if ZFS will be in Longhorn.
So can they not just collaborate with the really extensive OpenSolaris community to get this fixed then? Or when OpenSolaris is eventually released and available, will people just have to wait until a Sun engineer re-designs something?
Show us the filesystem or quit mentioning it.
😉
> Waiting for “Show me the Solaris or dont talk about it guy”!!
Oh, I’m here. And laughing.
Yes, got a problem buddy?
Yes. Where’s this Open Solaris open source community that was supposed to be involved in such things? If all they’re going to do is give previews of things through Solaris Express (which isn’t the same thing) then Open Solaris is essentially pointless. I look forward to the boy crying wolf again at the end of June, and having a look at just what happens.
We’ve got two much, much hyped, and not delivered, features that everyone at Sun was praising to the hills only for us to discover that they simply don’t exist. It is clear now that Janus certainly does not exist at all. Considering that these two features were supposedly a centre-piece of Solaris 10, and they’re not there, that basically means that Solaris 10 was never released.
And of Janus, he added, “We are about to release a technology preview of the Linux Application Environment for Solaris, with the goal of getting customer feedback on the best approach to this complex problem.”
Which means it doesn’t exist and never has done even when Schwartz and others were mouthing off about it, and they now want feedback on how to implement it?! Unbelievable.
Suppose a potential customer, however remote that chance is, had picked up on Schwartz’s and other Sun peoples’ comments and said “I want to run the stuff I’ve got on Red Hat Linux on Solaris 10 tomorrow. Where’s this Janus?” What on Earth would Sun have said? Do you think it would have convinced that customer to actually move from Red Hat, or someone else, to Solaris? Not a bit of it.
Sun really are full of a lot of the brown stuff at times, and I simply don’t see why because it isn’t necessary. They’re a company that lets their mouths run away with them without having thought things through properly.
First lets look at the announcement, so ZFS and Janus will not deliver on time. Am I dissapointed, YES! If I was going to configure a machine with Solaris 10 connected to several terabytes of data and use ZFS, I would like to know that it will work and work as advertized. I would rather have Sun pull something that doesn’t quite work than to include a feature that could trash data or hardware.
Unlike Microsoft, who never seems to learn their lesson about announcing new features and either not delivering on them or scaling them back, Sun (up until Solaris 10) doesn’t seem to say much at all. In a conversation I had with Chris Ratcliffe we discussed the problem of Solaris administrators not knowing about new features of Solaris and basically how to “get the word out”. Sun has added impressive functionality into their Solaris offerings, but unless you either stumble on the man page, run the command to see what it does, or hear about it from someone else, many have no idea the features exist. For years now it seemed to me that using Solaris was more like “digging for gold” to find those esoteric features. The Microsoft reference is nothing more than a cheap shot.
If ZFS does not work as it should, I applaud Sun’s decision to pull it back and get it right. The disadvantage of increased awareness of your products is potentially negative publicity when you “fail to deliver”. For the nay sayers and the Linux zealots, say whatever you like. I would prefer a company that errs on the side of caution than to release a product and back pedal when it doesn’t work.
Misleading title for this news blurb. It implies indefinitely while SUN has set CY06 as the new target date. I would hardly call that indefinitely. Not only that, the preview version of Janus should be available within the next month or so.
When i read the title, i thought they cancelled some major features.. is it FUD or what? Someone screwed up the title and has no guts to fix it..
Who really cares about Janus? Nobody in his/her right mind would ever deploy an app written for linux on solaris x86 using Janus. Its nothing more than a toy and Sun’s own engineers will tell you that off the record as one has told me that a few days ago.
The delay for ZFS is major though and even though I’m a huge Solaris fan and user I’m getting a bit tired or hearing them hype stuff that doesn’t exist yet.
If I was going to configure a machine with Solaris 10 connected to several terabytes of data and use ZFS, I would like to know that it will work and work as advertized. I would rather have Sun pull something that doesn’t quite work than to include a feature that could trash data or hardware.
Wow, really?
First lets look at the announcement, so ZFS and Janus will not deliver on time.
What do you mean “will not deliver on time?” They haven’t delivered on time. Schwartz and Sun were happy to shout from the rooftops about Janus and the virtues of running all of your current applications, and gave everyone the impression that it would be a part of Solaris 10, as they did with ZFS as well.
Unlike Microsoft, who never seems to learn their lesson about announcing new features and either not delivering on them or scaling them back, Sun (up until Solaris 10) doesn’t seem to say much at all.
Well, not much:
http://news.com.com/Suns+Solaris+10+to+run+Linux+apps,+too/2100-734…
(+ various other links)
Jonathan Schwartz, Sun’s former top software executive and now its chief operating officer, said he believes the Solaris-Linux technology will prove compelling to customers who want alternatives to Red Hat Linux, the dominant version of the open-source software and the one with the most compatibility certifications from software companies.
“In the data center,” Schwartz said, “your Linux vendor just tripled their price. You cannot move. Your application is not certified to Debian,” a Linux variant that hasn’t achieved mainstream commercial success. Solaris provides that escape hatch, he said.
A hand goes up in the audience. “Can I run all of my current stuff on Solaris tomorrow, and will Sun help move my stuff to Debian if I want it?” Errrrrrrrrrrrr, no.
Quite what that Debian reference had to do with anything I don’t know (are Sun doing Debian migrations these days?), but that’s the level that we’re working at here.
For the nay sayers and the Linux zealots, say whatever you like. I would prefer a company that errs on the side of caution than to release a product and back pedal when it doesn’t work.
Well you don’t like Sun then, because that’s exactly what they’ve done.
While it is great that Sun doesn’t want to put out crappy software, if they really embrased the open-source way, they could put it out half-done and work with more programmers to get it finished. Maybe someone could see a way of fixing the flaws without rewriting it from the ground up.
You mind citing a few more examples than the obvious references to ZFS and Janus. The basis for my comments about Microsoft go as far back as Windows 2000 (1999), where Microsoft proclaimed they were going to get rid of NetBIOS and WINS (which they still have in Windows Server 2003).
It is no secret that Janus is about using RedHat binaries on Solaris x86, so why should Sun help anyone port applications from Debian? Do you think IBM or HP is going to help you, I’m sure they will if you cut them a big enough check. So what’s the problem?
Out of all the new features in Solaris 10 (JDS3, SMF, DTrace, etc.), these were the only two important ones to get delayed. And they’ve actually been around for a while in beta. Outside of Longhorn/WinFS, Sun defintely comes in second place for the length of a beta program (at least recently).
The main people who are hit by the Janus delay are people looking to do server consolodation, which is generally something on a flexible time table (leave the current servers in place a while longer).
We’ve got two much, much hyped, and not delivered, features that everyone at Sun was praising to the hills only for us to discover that they simply don’t exist. It is clear now that Janus certainly does not exist at all. Considering that these two features were supposedly a centre-piece of Solaris 10, and they’re not there, that basically means that Solaris 10 was never released.
Agree with some of this. As someone has already pointed out,
Sun in the past has been criticised for keeping things new
in Solaris under it’s hat. Well, with 10 we decided to be
a bit more vociferous about some of the technology that
we are working on and what we believe will be differentiators between ourselves and the competition. With
engineering being engineering this has backfired a little
bit. However, I can state categorically that ZFS and
the LAE DO exist. ZFS is in testing (you must understand we
HAVE to get this right, peoples lives depend on it). The
first iteration of the LAE will be available imminently for
customers wishing to try it out. I think it’s fair to say
that Solaris 10 is still a landmark release even without
these two features. Opensolaris WILL happen and the Express
program is still a great way to see whats being integrated
into the next releases of Solaris. Oh, and it’s still free.
I read OSNews daily, and I hope OSNews would correct the title. It does not even match with its content. Sun customers, under NDA/CDA presentation, can know when the ZFS/LAE would make it. Unfortunately as a field person I’m not allowed to share to the public. There are many things we can share about ZFS/LAE, although sometimes sharing too much too soon might back fire. For readers who want to know more, just approach your local Sun reps, and you will get a lot more than what available on the Net, naturally. HTH. Iwan Rahabok, IT Architect.
Sorry to reply to myself, but another point is that Solaris 10 is zero-cost infinite right to use. It’s hard to complain much when I downloaded and installed it for the cost of 5 CD-R disks. And it is reliable like Solaris pretty much always is.
Considering that ZFS is the first 128-bit “self healing” filesystem going to ship in large volume to the world, Sun needs to get it right.
Janus appears to be more straight-forward (an emulation layer), so I wonder if it was just a lower priority feature relative to the other biggies.
”
Janus appears to be more straight-forward (an emulation layer), so I wonder if it was just a lower priority feature relative to the other biggies. ”
so they were pimping up a low priority feature big time?. thanks for clarifying
I’m just saying that Janus was probably slated to be released with the GA earlier this year, but prioritization probably caused it to be delayed. Talented engineers are a scare commodity, and if a lot are already working on other enormous features, not everything is going to pan out exactly as planned. IMO, it’s less likely to have show-stopper technical problems with something like Janus as with ZFS–perhaps they underestimated the volume of work to get all the details right?
That depends, if you are using Solaris x86 and intend to consolidate some RedHat machines, it is a big deal. I use primarily SPARC machines, so it doesn’t register as high with me as ZFS does.
Now you might see why I’m a Linux zealot.
I love being right.
“That depends, if you are using Solaris x86 and intend to consolidate some RedHat machines,”
so it is not a Linux compatibility layer, just a Red Hat compatibility layer?. that is not clear from their docs
ZFS delays are significant however who cares about Janus..name one Serious application that is Linux only that either doesn’t have a binary version for Solaris or another solution native to Solaris that could replace said application. Either way if the software isn’t ready it isn’t ready. There are already too many POS software apps out there that were released too soon.
What bothers me about Sun and J.S. and S.M. and MS is this typical capitalist hype machine.
When can we start competing by real features instead of these mythical pink elephants?
We’re still waiting. Compete! already, sheesh.
Former OS/2’ers here will probably agree with me that
Janus-ish features can be a mixed blessing.
If your support for somebody elses OS is too good
it might discourage companies from porting their
software to your platform -why should they pay the
extra development/qa/support costs for a Solaris
version when their linux binaries already work perfectly?
Note that I mentioned companies above. In particular
I am referring to companies/individuals who intend
to release closed-source (or pseudo-open source)
products where source code is unavailable or only
available under a difficult license.
I get the impression that a lot of open source Linux/*BSD
apps are quickly and “easily” (for a given value of
“easily”) ported to Solaris, so GPL/BSD licensed
stuff will probably rarely require Janus.
For this reason I think Janus is, ultimately, of questionable long-term value to Sun.
ZFS, on the other hand, is a real blunder on the part of Sun.
Nobody here, until now, has commented on the fact that
Eric Shrock is “completely redesigning the ZFS commands from the ground up” after finding some deficiencies.”
Leaving aside the question of how a talented Sun engineer could have made such serious errors in design that part of
ZFS must be redesigned from the ground up…
That leads me to wonder if ZFS might not be a very long time a-coming.
Yes, I realize that modern file systems are incredibly complex and demanding beasts.
>Leaving aside the question of how a talented Sun engineer
>could have made such serious errors in design that part of
>ZFS must be redesigned from the ground up…
You’re new to Sun, aren’t you?
Well if you can do any better, step up to the plate!
>Leaving aside the question of how a talented Sun engineer
>could have made such serious errors in design that part of
>ZFS must be redesigned from the ground up…
<Sharman>You’re new to Sun, aren’t you?
Yes and no. I’ve done QA work on Sun boxes, and
we’ve got about half a million dollars worth of
Sun hardware from about 4 years ago, but I’m not
privvy to the inner workings of Sun management and
tech support.
I have, however, followed the evolution of Linux JFS’es
fairly closely (to within the limits of my understanding),
and I do understand that there are some severely contradictory requirements (speed versus reliability, large vs small files, etc) that must be juggled very carefully.
The fact that Sun made significant errors in developing ZFS is not astounding – it could easily happen to a team of highly skilled engineers, and I’m sure Hans Reiser had his
share of bitter revelations too.
The problem is simply a matter of hype – you push an unready product too hard and you have to accept that people will come down on you like a ton of bricks if you badly fail to meet their expectations of when it will be ready.
skilled
Solaris is faster than Linux. But don’t try to benchmark it. You’ll void your license.
I wonder if they have similar clauses about mentioning bugs or other inconsistencies with the hype machine.
And just exactly where did you hear that from? Benchmarking Solaris will not void your license, however you might have to talk to Sun’s lawyers about violating the license agreement you “signed” by sharing that information with a third party without Sun’s permission. Read section 5(f) of this:
http://www.sun.com/software/solaris/licensing/sla.xml
Actually Sun encourages people to use the blogs, forums, and monitors USENET for people experiencing problems.
Seems as part of the deal they made with Microsoft, Sun has purchased a license from MS for vaporware.
“Sun Microsystems has delayed the rollout of the Sun Grid, an Internet-based “utility” service that has been under development since late last year. A lack of computing resources has pushed back the public launch, originally slated for the first few months of 2005, to as late as July, Sun executives said.”
Read more at:
http://www.pcworld.com/resource/article/0,aid,120678,pg,1,RSS,RSS,0…
Robert, I found your post a bit disingenious, and
possibly even a bit sinister.
What you are not saying is that Sun will likely start legal action against any individuals, web sites or magazines that dare to benchmark Solaris against some other operating sytem without Sun’s express permission.
[paranoia]
Your comments about Sun monitoring USENET makes me wonder if they aren’t actively looking for violations of this requirement.
[/paranoia]
Sun is playing a potentially dangerous game. The computer world is generally disenchanted with bogus benchmarks after years of shameless Microsoft spinning (remember MindCraft?).
Rigged and bogus benchmarks are a Microsoftism that Sun should definitely avoid emulating.
Given Microsoft’s history, why would anybody believe Solaris benchmarks that are authorized by Sun.
This applies to any sofware company, not just Sun and Microsoft.
Are you paranoid or what? It is Sun’s option (as I am sure any operating system vendors) not to allow benchmarks to be published without some review. Nowhere did I say that Sun is out to hunt people down. Sun employees regularly check comp.unix.solaris and alt.solaris.x86 looking for people with issues, and the last time I looked people like Casper Dik and Alan Coopersmith didn’t work for Sun’s Legal Department.
Let’s keep this professional and minimize the FUD shall we!
From the PC World article: “a number of recent large-scale grid deployments have forced Sun to divert systems that were to be used for the public site”
Sun is selling so many systems that they put their current customers first!
It’s amazing how taking things out of context works so well for trolls.
Robert:
Don’t you think it’s rather early to get accusative
and strident?
*You* were the one who pointed out that Sun monitors Usenet.
Are you now going to claim that Sun won’t take legal action against somebody who posts detailed, professional benchmarks in a Usenet forum, especially if the benchmarks make Solaris look bad?
if not, why did Sun put such a restriction in its license
agreement?
Early, let’s see you accuse me of posting “disingenious, and
possibly even a bit sinister” information. So let’s put this in context. Responding to the post by hmmm about benchmarks and licensing I made the comment that “however you might have to talk to Sun’s lawyers about violating the license agreement you “signed” by sharing that information with a third party without Sun’s permission. Read section 5(f) of this:” How does might become “will”, explain that to me.
Also my post never said anything about posting the results to Usenet. My comment was “Actually Sun encourages people to use the blogs, forums, and monitors USENET for people experiencing problems.” So while you are at it, explain to us how this consititutes the “Your comments about Sun monitoring USENET makes me wonder if they aren’t actively looking for violations of this requirement”.
How about not taking things out of context and twisting them into FUD. Sun monitors newsgroups for problems users and administrators are having with Sun products, so how is one connected to the other?
“Hmmm” claimed Sun didn’t allow benchmarks.
You corrected him, sort of, by pointing out that
Sun didn’t allow you to _publish_ benchmarks.
That’s the disingenious part. You are right, but
only on a technicality.
Nevertheless, I started contact with you on an inflammatory note, and for that I am sorry.
You see, I *want* people to be able to publish benchmarks
comparing Solaris to Linux to *BSD.
I want to read those benchmarks. I can’t perform valid benchmarks myself – I haven’t the skill or the hardware.
So I have to trust others, and frankly, I don’t trust Sun.
I very much doubt Sun will be anywhere near as shameless as Microsoft, but I won’t trust any benchmarks -from any company or organization – that prevents (hopefully unbiased) organizations or individuals from publishing their own benchmarks.
Seriously now – and this applies to everybody who bothers to read this post – wouldn’t you *much* rather see benchmarks from somebody who is as disinterested as possible, rather than someone who stands to gain directly from having one product look as good as possible?
as I am sure any operating system vendors) not to allow benchmarks to be published without some review
Red Hat
SuSE
Mandrake
Ubuntu
talk to Sun’s lawyers about violating the license agreement you “signed” by sharing that information with a third party without Sun’s permission
Is the opportunity to have such talks with Sun’s lawyers one of the benefits of the open community forming around the OpenSolaris source code?
This is a shame, as I’ve been looking forward to ZFS as well.
But to be clear what’s happening here, what is being “redesigned from the ground up” are the userland utilities that manipulate ZFS, rather than the ZFS core itself.
And the redesign seems to be around error reporting, rather than core functionality.
So, while the commands are being redone, it’s mostly a structural change rather than a logical or feature change. It sounds like they done all of the hard parts regarding interfacing to the ZFS kernel, etc., and are now focusing on error reporting and usability.
In the end, there will be a bunch of refactoring and cut-n-paste going on, so it’s not like he did rm -rf /zfs, and start from scratch.
A set back to be sure, but it’s not like they’ve been working on a 128-bit file system only to discover they only have 126-bits and need to go back and find the other two. Sounds like the core is sound, and they’re reworking the interface to it.
What is most telling, I think, is that Sun is letting him do it.
“What is most telling, I think, is that Sun is letting him do it.”
My bet is that Sun recognizes ZFS could become the next UFS, where it will be the standard in twenty years. Delaying one year for twenty of robust ZFS goodness is probably justified in their minds.
Everyone needs to recognize just how big a leap ZFS is. The 128 bits of addressing are a given, but ZFS is also supposed to provide self-correcting RAID, I presume along with early detection of failing drives, and they’re putting new volume management ideas on top of it. This is pretty serious stuff.
I appreciate your comment regarding ZFS and what it
could become.
Do you have any other information? Could you compare
ZFS to NTFS, Reiser4, XFS, JFS or EXT3? How
are other file systems lacking compared to ZFS?
And why didn’t you ask directly instead of this “cat and mouse” nonsense? Further, what hardware would you want these benchmarks run on, and what benchmarks?
The vast majority of benchmarks I don’t trust, regardless of who did them simply because there is so little material published as to what was done there is no way the test could be repeated and verified by an outside source.
Some of the research I have done indicates that the results you get for a particular system is only good for that system. You might or might not achieve similar results, and while it would be an interesting read, it could take months to complete. It also depends on what you are looking for, memory, disk, video, database?
But to be clear what’s happening here, what is being “redesigned from the ground up” are the userland utilities that manipulate ZFS, rather than the ZFS core itself.
Thank You…That is almost what I was getting ready to post after reading Eric Schrock’s blog:
http://blogs.sun.com/roller/page/eschrock/20050414#designing_for_er…
As I said before, let’s keep the FUD down to a minimum. OpenSolaris is a totally different product, the topic of discussion is ZFS, which is part of Solaris 10.
“OpenSolaris is a totally different product, the topic of discussion is ZFS, which is part of Solaris 10.”
OpenSolaris is not a totally different product. it it supposed to be the open source version of solaris managed by the same company. If they poorly execute one thing it affects the other
ZFS is not a part of solaris 10 either. There is no current date being set for its release. read the above blog
The linux compatibility/emulation layer only works on x86..not sparc. This is as per the engineer who gave a Solaris 10 preview that I attended. It makes sense if you consider cpu architecture.
Also, he mentioned that Sun will consider it a BUG if a Linux app runs slower on Solaris x86 emulation than in the native linux environment. That’s pretty ballsy.
i don’ t care about janus but i’ m sad for zfs that sound promising and interesting.
but wen is supposed to ship opensolaris?
As if you were ready to use ZFS right now. It’s very advanced as far as FS’s go, but you need to be an enterpise to be able to benefit, I think. Enjoy.
You mind citing a few more examples than the obvious references to ZFS and Janus.
Apart from Sun emplyees, and execs, like Jonathan Scwartz actually saying those things you can’t go much further.
It is no secret that Janus is about using RedHat binaries on Solaris x86, so why should Sun help anyone port applications from Debian?
Because Schwartz gave that as a specific example? When people are sick of Sun and Solaris are Sun going to help people move to Debian?
“And why didn’t you ask directly instead of this “cat and mouse” nonsense?”
First off, I wasn’t playing “cat and mouse”, whatever that
is, and I don’t believe I was engaging in any nonsense.
Secondly, I’m not sure what I was supposed to “ask directly”. I didn’t have any question for you – I just
didn’t like the fact that you were, as I said, being
disingenious.
The topic, as I recall, was benchmarks.
“Further, what hardware would you want these benchmarks run on, and what benchmarks?”
You probably didn’t intend to do so, but you’re playing straight man to my soapbox here…
I want lots of benchmarks – I want a wide variety of the most popular, well-supported hardware to be tested. I want benchmarks for file serving, for web serving, for database sorts and searches, I want frame rates (where relevant).
Sun (and sun-approved/financed organizations) couldn’t produce as many benchmarks as I’d like to see, because
(as you said):
“while it would be an interesting read, it could take months to complete.”
Good, reproducible benchmarks *do* take a lot of time and effort, and Sun (and its approved benchmarkers) wouldn’t
have time to do a wide variety.
“The vast majority of benchmarks I don’t trust, regardless of who did them simply because there is so little material published as to what was done there is no way the test could be repeated and verified by an outside source.”
I agree that some benchmarks are described too vaguely to be reproduced. Such benchmarks are of little value.
I’m not so sure about “the vast majority” bit though.
Also, I’ve seen benchmarks which use custom-tailored drivers that were not available to anybody outside of Microsoft. These benchmarks are also useless in the real world.
“Some of the research I have done indicates that the results you get for a particular system is only good for that system.”
I’m not sure how to interpret that comment – it is so bizarre I can’t believe you’re saying what I think you’re
saying.
I agree that results on, say, an HP PIII uniprocessor with an 8 gigIDE drive won’t necessarily scale consistently to
a quad Xeon with RAID0 160 megabit SCSI drives, but my
2.4 gig Dell with Seagate 120 Gig hard drive should perform the same as the tester’s Dell with identical hardware and identical software configuration.
Again, that is why I would love to see lots of benchmarks -many different tests on a wide variety of hardware would highlight strengths and weaknesses of the OS’es being tested.
“You might or might not achieve similar results, and while it would be an interesting read, it could take months to complete.”
That’s why you need to have many credible people doing the benchmarks.
With engineering being engineering this has backfired a little bit. However, I can state categorically that ZFS and
the LAE DO exist.
I don’t doubt that they do exist in some form, but there’s a difference between that and releasing something. The LAE (Janus etc.) has simply been used as an obvious knee-jerk reaction to Linux, and Red Hat in particular, and that doesn’t help. You’ve then got execs mouthing off stupidities about those things, and mark my words, the customers you want to attract and capture see that as stupidity.
You want to know the secret as to why Sun get a lot of vitriolic, intense and sometimes troll comments? Because you’re an infuriating company; you cheese people off.
You’ve got (and have had more in the past) all the ingredients to be absolutely wildly successful. Linux is obviously a large stone in your shoe because you want to be protectionist with Solaris, and it shouldn’t be. If you could rub two brain cells together you could be a much better Linux vendor then either Red Hat or Novell (no agreement with Microsoft necessary – quite the opposite). But I don’t think you can because you just don’t grok it.
ZFS is going to be very cool, and not just for big enterprises. I think its going to be a wonderful asset to anyone who has more than one disk drive, or who has ever made a partition too small. (Such as yours truly that has a pretty mangled UFS volume at home…)
Basically, ZFS is to offer the basics of any modern filesystem, but it also includes features like raiding, striping, and snapshots (mmm…snapshots). It also makes it easier to mix and match physical volumes into logical volumes.
So, is your /usr partition too small? Slap in a new physical drive and seemlessly extend it then. Things like this are supposed to be very easy to do.
While one of the focuses is to make it easier to manage insane numbers of drives, a lot of that management will boil down into making it easy to manage two drives as well.
The key is not that these are new and innovative features, they’ve been around on other systems and tools for some time, but the fact that Sun basically decided that the world has learned a lot about file systems in the past 20 years, it’s time to redo from start and remove as many limits as practical from the integrated package. On top of that, they’re giving it away (there is little doubt that ZFS will fold in to OpenSolaris at some point).
Veritas, Netapp, EMC, etc. have to do some major marketing to convince the market that their feature set is better than the free download from Sun.
Check out: http://www.sun.com/2004-0914/feature/
I spent fifteen years doing photographic quality control for the US Navy prior to getting into IT. When you test a photographic system (camera, processing machine, etc.), the results tend to be specific to that system. Benchmarking computers is the same, even with the exact same components, the results will be similar but not the same. The results are specific for that machine, that OS, and that test. At best it is a generalization of how something will respond based on a given set of criteria.
Operating system benchmarks are some of the most complex, because to tune for one result skews another (if they were going to run the tests consecutively). And this is where the problems start, because most vendors don’t like “out of the box” results (neither do I). So they tune, in some cases the tuning steps are made public, in other cases they are not.
I just don’t see the point, each OS has strengths and weaknesses. Running a benchmark like iozone is pretty straightforward (www.iozone.org), I do it all the time. What I recommend is you get some hardware and test for yourself. I found it to be an interesting experience.
And why would Sun want to be another “Linux” company when they have a better product (Solaris)? Just because IBM and HP did it doesn’t mean Sun has to do it.
And why would Sun want to be another “Linux” company when they have a better product (Solaris)? Just because IBM and HP did it doesn’t mean Sun has to do it.
Some people believe they have the better product, some don’t, but that doesn’t matter because they’ve been getting their asses kicked for several years and now they’ve got the faint hope of some people coming back. Besides, Solaris doesn’t matter. What’s more interesting is the stuff they do on top of Solaris and that’s what could make the real difference for them
Sadly, they just don’t get that.
How about specifying, I’m curious.
>And why would Sun want to be another “Linux” company when
>they have a better product (Solaris)? Just because IBM
>and HP did it doesn’t mean Sun has to do it.
That’s a tall claim… and a demonstrably false one in many cases. Better how? Can I run my Winmodem and voice-mail capability on Solaris? How about my iSCSI and Infiniband cards? What about my Broadcom GigEth? No?
Do a little research… Solaris has its strong points. So does DOS.
> Can I run my Winmodem and voice-mail capability on Solaris?
Well you can’t run it on Linux either – so no cigar for Linux.
> How about my iSCSI and Infiniband cards?
Abso-f***en-lutely!
http://www.sun.com/io_technologies/infiniband/infin-index.html
> What about my Broadcom GigEth? No?
http://www.broadcom.com/drivers/driver-sla.php?driver=570x-Solaris
>> How about my iSCSI and Infiniband cards?
> Abso-f***en-lutely!
> http://www.sun.com/io_technologies/infiniband/infin-index.html
In fairness, that link is actually from IHV meetings that took place five years ago… so definitely not the best, most modern pointer. But, nonetheless, Solaris 10 does absolutely support Infiniband. (And Solaris Express today even provides the additional support for the modern PCI-Express flavors of IB cards.)
Here’s a good place to start for more modern info. But search around, there’s plenty of info online:
http://docs.sun.com/app/docs/doc/817-5093/6mkisoq0c?q=infiniband&a=…
You mean these drivers:
http://www.broadcom.com/drivers/downloaddrivers.php
Look for Solaris x86
Maybe you need to do some research before spouting off!
You will notice they only have drivers for some broadcom chipsets here:
http://www.broadcom.com/drivers/downloaddrivers.php
Just like Linux has support for some Winmodem chipsets here:
http://linmodems.org/
So, from this perspective Solaris is almost as good as Linux.
From a licensing perspective, however, Solaris still sucks. And it looks like it always will. This is why we were pushing for OpenSolaris. But does it really matter? We already got Linux and BSD. Sun doesn’t have to release Solaris under the GPL. They really don’t. Its just not important anymore. A couple years ago, when I first mentioned it, it might have given them some advantage. But now? Well, 2.6 is stable. Its kinda late now.
Shaman wanted Gigabit Ethernet drivers, well according to what I read Broadcom 57xx drivers are available for Solaris x86, Linux, and even for SCO Unix. And who in the hell would use a Winmodem? Why would Sun or the vendor of a Winmodem even bother writing drivers for Solaris?
If you are going to complain about hardware support, let’s be realistic. If I remember correctly, even Linux had problems with Winmodems. I could very easily come up with some piece of hardware that one in a thousand people would use and mindlessly bitch about no support for it with a particular OS. Not every OS, not even Linux is a panacea of hardware support.
> You will notice they only have drivers for some broadcom chipsets here:
What do you mean?. It’s a broadcom.com site – you don’t expect to see intel gigEth drivers there, now do you?
You want intel chipsets for Solaris x86? (damn man do I have to google again?): http://downloadfinder.intel.com/scripts-df-external/Detail_Desc.asp…
@yawn
[i]From a licensing perspective, however, Solaris still sucks.</i.
Really? That’s funny, my company thinks it’s great. Guess the real business folk and developers will use the “suck” thing you’re deriding in glee while you’re still sitting in your armchair criticizing things you don’t understand.
Compared to the GPL?
All I see are idiots who state that Linux is superior to Solaris because it has more drivers, and then it turns out that those claims are false because Solaris has those drivers.
And idiots who criticize Solaris (Ipresume OpenSolaris?) because it is/will not be released under the GPL, while praising Linuxand BSD…. while BSD is obviously not released under the GPL either.
Or idiots criticizing Sun for sticking with it’s proprietary UNIX unlike IBM and HP, while everyone with a 1/4 of a brain knows that IBM and HP are sticking to their proprietary versions of UNIX (AIX and HP-UX), and have no intention to open source either of them anytime soon.
A forum that is supposed to bring bright minds together, actually attracts some stupid people.
It’s like being in high school all over again. The people complaining the most are those most driven by fashion and the appearance of popularity, and they allow others to form their opinions for them. The facts are that Solaris 10 licensing is awesome (unprecedented for UNIX), and Solaris 10 itself is a huge leap from Solaris 9 (the biggest jump since Solaris 7, IMO).
Thank you! I went round and round with a troll over the very point of where’s IBM and HP in open sourcing their operating systems. His argument was “where’s the value for IBM”, ignoring the obvious increased exposure could potentially bolster IBM’s bottom line. Especially for those who think using a *nix product is “too hard”. I could take somebody off the street and get them to admin an AIX machine using smit in a few minutes.
I think many of these people are burying their heads in the sand and ignoring commentary when anybody has the intestinal fortitude to mention the possibility that IBM and HP have alterior motives behind their support for “OSS”. The next few months will be interesting to see what the response by IBM and HP is when OpenSolaris rolls out.
It will be interesting to see what response (if any) we get from these posts.
“The next few months will be interesting to see what the response by IBM and HP is when OpenSolaris rolls out.”
They won’t know how to respond. OpenSolaris will be the best-of-breed open source operating system. It will allow ISVs and customers to better integrate and troubleshoot their products and projects. Hardware vendors will get a consistent and open API to write drivers to. It will allow hobbiests to learn how software backed by billions in R&D works. It will benefit from the fact that Sun isn’t a fly-by-night Sourceforge project. It will remain consistent and documented under Sun’s management. It will be the same operating system running the largest companies in the world. It largely renders AIX and HPUX obselete due to cost, especially teamed with Linux, which is also free.
Sun was working on OpenSolaris for quite a while before they announced it, giving them a few years head start, which strategically was a genius move against IBM and HP. It could even ultmately be the undoing of Microsoft, but that is a bigger target with a longer timeline.
idiots who state that Linux is superior to Solaris because it has more drivers, and then it turns out that those claims are false because Solaris has those drivers.
http://www.broadcom.com/drivers/downloaddrivers.php
Where are the bcm4401 Solaris drivers?
idiots who criticize Solaris (Ipresume OpenSolaris?) because it is/will not be released under the GPL, while praising Linuxand BSD
At least I can get the source code to Linux and BSD. Let me know when you finally build your first OpenSolaris system from source.
idiots criticizing Sun for sticking with it’s proprietary UNIX unlike IBM and HP, while everyone with a 1/4 of a brain knows that IBM and HP are sticking to their proprietary versions of UNIX (AIX and HP-UX), and have no intention to open source either of them anytime soon.
Neither IBM nor HP claimed they were going to open AIX or HPUX respectively. If Sun wants to remain proprietary you will hear no complaints from me. I thought, 2 years ago, it would have given them an advantage over Linux to open source Solaris with the GPL. But even if they do open the Solaris source code with the CDDL today it still won’t make a difference. Its too late. It can’t compete with the momentum of Linux. It takes years to build up a community like that, these things won’t happen the day OpenSolaris is released. Even the BSDs have a hard time keeping up with Linux. No matter what Sun does at this point it won’t affect the adoption or development of Linux. Even going so far as to GPL OpenSolaris today would have little effect.
But go ahead, call me an idiot.
“It can’t compete with the momentum of Linux. It takes years to build up a community like that, these things won’t happen the day OpenSolaris is released.”
Solaris already has tons of momentum in a huge installed base. The community is already there. The pilot program has very enthusiastic people in it. Opening the source code will only push this further.
ISVs would _love_ OpenSolaris. It’s APIs are consistent, QA against new releases is easy, Solaris is very well documented for newbies and old farts alike, and it is generally a pretty clean system. Even in Solaris 10, with a gutted and replaced init system, compatibility is still retained.
People who criticize the CDDL don’t understand just how easy it is for ISVs to play along. The full spectrum of ISVs are accomodated, from the most OSS-friendly ones to ones with trade secrets to ones so old-fashioned OSS is forbidden. They can all release drivers for OpenSolaris, they can all derive OpenSolaris for their own needs. They can even resell it with no royalties to Sun.
Cute name.
They can all release drivers for OpenSolaris, they can all derive OpenSolaris for their own needs. They can even resell it with no royalties to Sun.
They can do this today?
They can do this today with Linux and BSD. So what’s your point?
”
Solaris already has tons of momentum in a huge installed base. The community is already there. The pilot program has very enthusiastic people in it. Opening the source code will only push this further.”
if solaris had the momentum already SUN would never even think of opensourcing it. SUN is losing heavily due to Linux and this is their idea of striking back…
”
ISVs would _love_ OpenSolaris. It’s APIs are consistent, QA against new releases is easy, Solaris is very well documented for newbies and old farts alike, and it is generally a pretty clean system. Even in Solaris 10, with a gutted and replaced init system, compatibility is still retained. ”
you mean like oracle switching over to Red Hat or Novell who have given ISV’s same consistent ABI. Kernel internal API is none of ISV’s business. The above layers and things like the GNOME fork called “Java” Desktop system which is NOT ABI compatible with the upstream version. SUN finally managed to ditch their proprietary X fork for Xorg which is pretty much the same everyone in Linux and BSD. are you talking about samba?. what other sub systems are ABI compatible that Linux vendors dont ?
“People who criticize the CDDL don’t understand just how easy it is for ISVs to play along. ”
you mean for those open source people to feed in code to “open” solaris so that ISV’s can take advantage?
“The full spectrum of ISVs are accomodated, from the most OSS-friendly ones to ones with trade secrets to ones so old-fashioned OSS is forbidden. ”
you mean proprietary vendors would be eager to develop proprietary kernel modules for SUN?. that can be done for BSD too. whats the idea here?
And what exactly is the point to having the source code? Explain to us how this is an advantage over any OS? I see it as a potential vulnerability, just as the leak of Microsoft’s Internet Explorer souce code yielded a number of difficult to eradicate exploits, and the leak of Cisco’s IOS code could produce new exploits, Linux is one zero-day exploit from being rooted en-masse. The serious criminals and malicious users are just waiting for market penetration in the right sectors, and if you don’t think it can happen, think again! Yes it sounds paranoid, but good security people are a little paranoid.
Closed source is not perfect, but it usually requires a great deal more effort to hack than to hand the source code over to anyone who wants it. So what is the benefits of having the source code again? If you are not a developer, an engineer, or a security analyst, the source code is useless.
Sun has had over twenty years to penetrate the market, the same as IBM and HP. In the area of the country I live in the vast majority of installed Unix servers are Sun, HP comes in second, and IBM a distant third. And with all of the woes HP is experiencing at the moment, gaining momentum is something they need to do if they are going to continue to be an enterprise player. Unless they want to become another Dell and sell Windows servers.
The “momentum” of Linux is based on people who are interested in using it and having it meet their needs. Community means nothing if nobody wants what your “community” is developing. It comes down to what can Linux do that Solaris (or AIX or HP-UX) can’t, and if you have a large base of Sun< IBM, or HP hardware the decision is obvious (if you have a clue). Linux is not a panacea of software support either.
And once OpenSolaris is released, it will get reviewed by security analysts and Black Hats for vulnerabilities that could be exploited against Solaris 10 installations. Does this necessarily help improve security, I think the jury is still out on that.
> At least I can get the source code to Linux and BSD. Let me know when you finally build your first OpenSolaris system from source.
Solaris doesn’t suffer from Linuxitis and hence you don’t need recompiling drivers when you patch the Solaris kernel. Solaris , like BSD and unlisk Linux has a stable DDI. Linus purposely designed Linux not to have a stable DDI because how else would you control binary modules taking down the kernel. So rather than have a stable driver interface, he chose having a stable user-level interface and change drivers accordingly. It’s just wrong!!!!. if an app has a problem fix the app not the driver…..it’s like my Hummer SUV doesn’t fit within the lanes on the freeway so the govt. should repave the freeway so that my SUV fits.
Linux does have a fairly stable binary interface but the kernels programmers do NOT want it to be stable. Why? Because they believe that there should be no proprietary binary drivers in use with linux. They should be contributed in source form to the kernel managers.
“Because they believe that there should be no proprietary binary drivers in use with linux.”
Artificially creating barriers to entry is just great for gaining credibility among hardware vendors and potential customers. That is if gaining means losing.
Sun will always have the advantage of actually giving people what they want. The hard-line Free Software folks will end up being their own downfall, because telling other people how to live their lives just pisses them off.
Good quality company line answer!
… [Linux] kernels programmers do NOT want [the binary driver interface] to be stable.
…they believe that there should be no proprietary binary drivers in use with linux.
FUD
The kernel programmers want a stable interface, but they also want to keep full control of it. This means you should not expect it to be static, it can and will be changed as necessary.
They do discourage binary drivers because it means you would have to make any changes necessary to your drivers to maintain compatibility with the kernel’s binary driver interface as it changes. Also, it is much harder to fix a stability or security problem in a closed proprietary driver. And if your driver is GPL and integrated into the kernel it will be automagicly updated by the maintainers with any changes made to the module/driver interface. And in additional to all this you would be able to use subroutines and code from anywhere in the entire kernel or userland GPL source that’s available everywhere. That’s millions of lines of functions and methods and data and good quality code.
Linux does have a fairly stable binary interface
Yep. So it is possible to maintain your own proprietary or combination of open source and closed binary drivers like ATI and nVidia. That way you can still take advantage of the community to keep the glue between your driver and the kernel stable.
The problem I see here is a lot of people feel threatened by open source. I do too. The company I work for has patents and code that they could not open source without it threatening their business. But that’s not what this whole Linux thing is about. Its about building a stable platform to use for these proprietary patented kick ass applications. Its about getting and giving you full control of your hardware.
What would you rather have? A stable binary driver interface, or an OS that won’t bow down to the demands of governments or media companies and offer you everything it can technically achieve in your lifetime for free?
I want more than stability. I want unlimited access to all available improvements for free. I demand more. Now.. gimme!
Since you agree with me, I’m struggling to understand how what I said was FUD. The kernel developers do, in fact, regularly break proprietary drivers (think ATI and nVidia) with changes to the kernel without any apologies.
And they want it that way. Because those drivers do taint the kernel and an inattentive vendor will cause problems for end-users by not updating drivers. What they want, as you point out, is vendors to put their code in the kernel tree.
I’ll never entirely understand why hardware vendors wouldn’t want the kernel developers to do the majority of their work for them. Because that’s what they get, if they provide specs and example code to the kernel devs.
Oh so all that marketing hype was for nothing? The touting of ZFS as the new end all FS that will redefines hard drives as we KNOW THEM AND BEYOND! Hmmm now that I think about this, this is very bad since we will have to hear all the hype again when ZFS is actually shipped.
> I’ll never entirely understand why hardware vendors wouldn’t want the kernel developers to do the majority of their work for them. Because that’s what they get, if they provide specs and example code to the kernel devs.
Look Linus Torvalds or any of the kernel developers know SHIT about GPUs or OpenGL or anything graphics related. They are excellant knowing how x86 and other CPUs work. But trying to write a driver to get the best performance from a GPU is best left to the guys who designed the chips!.
ATI and Matrox gave tons of docs to the X guys, but the X drivers are nowhere close to the quality/feature of Windows graphics support or even support from XiG.
On a ATI Rage (supposedly 100% open source) Quake2 gave be horrible frame rates with MesaGL – While Windows on the same card with ATI’s rage/GL drivers works just fine.
Look, people are good at open source development aren’t necessarily good at developing every thing. Like you go to your general medical doctor for normal medical ailments but you’d never trust your personal doctor with a heart bypass surgery. Same way, I trust kernel developers to write good general purpose code but given the option between a closed source Nvidia driver and an open source implementation (even if Nvidia gave the specs) I’d rather trust Nvidia’s implementation any day.
This basically boils down to who would you trust?. A kernel from Linus’ kernel tree or a kernel from Shaman’s tree?
What do you think about the Sun – Microsoft announcement?
I hope it isn’t another attempt to shove something like PKI down our throats again.
MesaGL doesn’t have any hardware smarts to speak of. That’s why it is slow.
Most of the drivers in the kernel are done from specs including the CPUs. If a vendor wants to make it work best, then obviously they should be involved in development. Then the kernel devs and programming communities involved can help debug (and they will).
Oh so all that marketing hype was for nothing? The touting of ZFS as the new end all FS that will redefines hard drives as we KNOW THEM AND BEYOND! Hmmm now that I think about this, this is very bad since we will have to hear all the hype again when ZFS is actually shipped.
How is that any difference than the linux guys touting the O(1) scheduler for the two and a half years the “accelerated” 2.6 development took place? Or the per cpu dispatch queues? or 1:1 thread library?
All of those Solaris had for a decade before linux developers even thought about putting it in. Every newspaper and media was touting these things as linux technologies that would revolutionize the world.
Delaying a couple of features till they are perfect is nothing new to linux. Why is it a problem for Sun and Solaris in the eyes of linux zealots. Double standards? 2.6 still isn’t considered stable. Right?
“Double standards?”
Absolutely. People just like to bitch about Sun. The CDDL is MPL-derived and the APSL is MPL-derived, but people just salivate over Apple. The Linux kernel tech is another example, where Solaris has had this stuff for years.
It’s just a fact that Linux and Apple have vocal fan clubs, who don’t mind putting fashion and hype over facts.