Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. Also, Slashdot's Facebook page has a chat bot now. Message it for stories and more. ×
Sun Microsystems

Sun Works to Converge Linux and Solaris 291

Jucius Maximus writes "A new change has appeared in Sun's strategy as reported by CNET. Instead of dismissing Linux as inferior, it is now trying to integrate elements of Linux into Solaris for easier porting of applications. This looks like a step in the right direction for Linux acceptance in the professional server market."
This discussion has been archived. No new comments can be posted.

Sun Works to Converge Linux and Solaris

Comments Filter:
  • GPL (Score:3, Insightful)

    by Andrewkov ( 140579 ) on Tuesday May 14, 2002 @11:59AM (#3518065)
    I'm sure Sun's lawyers have been over this with a fine tooth comb, but if they integrate GPL code fron Linux into their OS, doesn't that mean they have to release the source for their whole kernel? Or is it just libraries?
    • Re:GPL (Score:5, Insightful)

      by the red pen ( 3138 ) on Tuesday May 14, 2002 @12:05PM (#3518116)
      Sun is making sure that Linux API's are available on Solaris. It is certainly possible to do this without having to GPL all or any of the code base.

      They are doing this so that Linux applications will port trivially to Solaris. This isn't an admission that Linux is as good as Solaris, but just that it's more popular for developers.

    • So long as they don't copy the code they are fine. It is well understood that you can look at GPL code to get ideas for how to implement things, then use the abstracted ideas to write your own implementation. Just don't cut and paste, or end up with code that makes it look like you did.

    • From my understanding, if you only link drivers into the kernel, you do not have to GPL them. Only changes to the kernel need to be GPL'd as well. I also believe that Sun's intention is to integrate more GNU applications, not necessarily kernel code.
    • Reverse Strategy (Score:5, Interesting)

      by 4of12 ( 97621 ) on Tuesday May 14, 2002 @12:18PM (#3518233) Homepage Journal

      if they integrate GPL code fron Linux into their OS

      Instead of integrating Linux code into Solaris, what Sun needs to do, rather, is to implement some of the nicer features and interfaces of Solaris into the Linux kernel, making Linux look more like Solaris.

      I mean, it already does in a lot of ways and, to be sure, they'll have to contend with differences of opinion from the benevolent dictators that control the Linux kernel (eg, POSIX threads debate), glibc, etc.

      But it's in Sun's best interest to pave a smooth superhighway upgrade path from Linux to Solaris for users that grow beyond their x86 hardware.

      Also, with their ownership of Cobalt, they could really make a pressing low end solution of Java on Linux/x86 to build flavored servers using open source interfaces without tying clients into a OurOneSizeFitsAllYourNeeds scheme. Then, customers wanting more complex business logic could opt for slicker building environment that Sun could sell.

      The other hardware route that Sun could take is to build an x86 system with the hardware reliability that has been lacking, especially compared to SPARC systems. Linux gives you a UNIX OS with plenty of nines, there's no excuse for the hardware to crap out as much as it does, especially for servers.

      • by Andrewkov ( 140579 ) on Tuesday May 14, 2002 @12:36PM (#3518364)
        But it's in Sun's best interest to pave a smooth superhighway upgrade path from Linux to Solaris for users that grow beyond their x86 hardware

        Well said, I think this is the reasoning behind the move. Let "newbies" migrate their servers from Windows to Linux (hey, it's cheap), then once they outgrow their x86 box it will be easy to move to Sun hardware without porting their business application or whatever it is their running. Brilliant idea! I guess Sun realises it is better to have people using Linux than Windows, since they are more likely to move to Sun from Linux than they are from Windows. Linux makes a good intermediate step.

      • Instead of integrating Linux code into Solaris, what Sun needs to do, rather, is to implement some of the nicer features and interfaces of Solaris into the Linux kernel, making Linux look more like Solaris. ... But it's in Sun's best interest to pave a smooth superhighway upgrade path from Linux to Solaris for users that grow beyond their x86 hardware.


        How would adding cool Solaris features to Linux help Sun:
        1. migrate more users from Linux to Solaris? Why buy the cow when the milk is free?
        2. sell more Sparc servers? The Solaris features added to Linux would be GPL and quickly ported to x86 Linux, undercutting Sparc server sales.
        • by 4of12 ( 97621 ) on Tuesday May 14, 2002 @04:00PM (#3519929) Homepage Journal

          How would adding cool Solaris features to Linux help Sun:
          1. migrate more users from Linux to Solaris? Why buy the cow when the milk is free?
          2. sell more Sparc servers? The Solaris features added to Linux would be GPL and quickly ported to x86 Linux, undercutting Sparc server sales.


          Good questions, and I'm not sure the answers are as comforting as you would like. But here goes.

          It wouldn't actively cause migration from Linux to Solaris, but when the time comes to upgrade from a low cost Linux solution to something bigger and better, will it be Sun, HPaq, IBM, SGI? If Solaris is most compatible, then that choice will be Sun. Of course users will drink free milk as long as they can, but someday they'll need more milk than the free cows can give. The key is to insure that low end cheap server market goes into the UNIX world on a upgrade path that leads to you rather than your competitors.

          Low end SPARC hardware sales are a losing proposition at this point in time. Sun has reasonably good high end offerings, but in the low end they're offering Solaris/SPARC vs either Wintel or Lintel. Lintel is the ultimate lowest cost option and, while it eats the lunch of low- end SPARC, that lunch was going to be eaten anyway by either Wintel or Lintel. At least the Linux box keeps users in the UNIX world where Sun has a lot of software experience to offer. Logically, you want the cost of rewriting business application logic transitioning from Linux to Solaris to be small. Also, 64-way Solaris/SPARC machines give those UNIX users at the low end a scalable upgrade path that is missing from the Wintel world.

          Granted, it's harder to make money where much of what was previously offered is becoming commoditized, but it's an irresistable force that the market is demanding.

          Ride that wave and anticipate where it's going instead of trying to stop it.

          • Re:Reverse Strategy (Score:3, Interesting)

            by ahde ( 95143 )
            Actually, unless we're different than most, the Sun Netras (at $1000) was near the top of the heap when I went looking for 1U rack servers. Way ahead of the RAQ4, and anything from any of the major x86 players. Sun's support and Solaris' compatibility with enough Linux apps almost won out. If it shipped with gcc or there were more packages out there for Linux on Sparc, we might have gone with the Netra.
    • What exactly is it about SysV cp, mv, tar, awk, ls et al that makes them so much more valuable than their GNU equivalents?

      Sun has no idea how to address Linux. However, if Sun were to replace all possible SysV components in Solaris with their GNU equivalents, they would be much farther down the road towards a free OS than the Sun Community Source License ever got them. This would at least give them some short-term PR, plus cutting development costs.

      I really don't understand why every UNIX distribution isn't making these moves. If I were to say that 90% of the GNU UNIX utilities could replace the proprietary components with no visible effect to the OS, would that be a conservtative estimate?

      Sun could go even further by wrapping Red Hat Linux around the Solaris kernel, and scaling Red Hat onto an e15k.

      And, if Sun were to take the step of open-sourcing the Solaris kernel, Sun could put an end to the question of enterprise UNIX on any Intel platform - Sun takes all.

      Come on, guys, wake up! You're asleep at the wheel!

      • by irix ( 22687 ) on Tuesday May 14, 2002 @02:01PM (#3519040) Journal
        if Sun were to replace all possible SysV components in Solaris with their GNU equivalents

        It would cause chaos. Come on, they just can't ship Solaris 9 and replace the Sun tar with the GNU tar. I'll give you that GNU tar is way better than the one Sun ships (the GNU tar comes first in my $PATH), but people have written software (Solaris package install scripts, for example) expecting the Sun tar to be there and take a certain set of arguments. Maybe tar is a bad example, but you get the idea.

        Sun is doing the smart thing by gradually switching things over. They have some GNU stuff available in the core install, and some GNU stuff available on a second CD [sun.com]. I now can write software for Solaris assuming Perl is installed, for example.

        This will improve slowly, over time.

      • I really don't understand why every UNIX distribution isn't making these moves.

        Because the GNU utilities also come with Richard Stallman hounding you and everyone connected with you to rename your OS and prefix a GNU/ in front of it...

        More seriously, they come with the GPL: while 'cp' isn't exactly primo intellectual property, it makes even the developers (not just the lawyers) happy to have a base system that's all owned -- coders would rather think about technical compatibility when they copy or link in a routine from another subsystem ... not political compatibility. Developers don't want to lose their job because the flagship product needs a critical component replaced if it doesn't want to become GPL'd.
  • by TuxLuvr ( 578149 ) on Tuesday May 14, 2002 @12:00PM (#3518071) Homepage
    This is a refreshing change of direction for Sun, but it may be too late in some instances -- as we all know, many companies are now phasing out their Sun hardware for inexpensive Linux-based solutions...

    What I would like is for FreeBSD to include Sun binary compatibility in 5.0, so I can run my Linux apps inside a Solaris VM under FreeBSD! ;^)

    • Hu?

      Why should I phase out old sun hardware for cheap intel based hard ware? Linux runns pretty well on SPARCs.

      (Not to mention that most SPARCs I work on run SUN OS 4.3 and not Solaris)

      angel'o'sphere
      • Why should I phase out old sun hardware for cheap intel based hard ware?

        Fair enough -- but Sun support contracts can really hit the ol' austerity budget hard, and the cost of renewal may not be worth it for some...depends on the situation of course.

    • This is a refreshing change of direction for Sun, but it may be too late in some instances -- as we all know, many companies are now phasing out their Sun hardware for inexpensive Linux-based solutions

      But in the end, Solaris is still better than Linux on big iron, and there's more margin in big iron. While SPARC is on the way out for the workstation market, I think it'll be around for a long time for $20k+ servers, and so will solaris.
    • from the article:
      In a dramatic departure in February, though, the company embraced Linux as well, with Chief Executive Scott McNealy dressing as Tux, the penguin mascot of the Linux movement.
      Surely, someone can link a picture of this, right? Please?
  • by dingo ( 91227 ) <(ua.moc.tentsew) (ta) (sdrawdeg)> on Tuesday May 14, 2002 @12:01PM (#3518077) Journal
    Sun is trying to put Solaris features into linux.
    This strikes me as a very bad move. Why would you improve "competing" products. Now addmitedly it will help them sell more solaris machines but given the open source nature of linux wouldnt this mean the improvements could be relatively ported to intel...effectively shooting themselves in the foot.
    • Becuase Sun could stand to benifit and possibly profit from a version of linux with some of thier well engineered and as of now, closed source OS code. How do you make money with Linux you ask? By selling support for the product. RedHat has been sucessfully doing this for years. Imagine how much better an enterprise version of Linux would sell if it had the Sun stamp of approval on it. Sign me up for beta testing. I have always wondered how the VM on Solaris worked anyway
    • Not really, in my opinion. What Sun hopes to gain is two systems which are more or less completely compatible. So, if you want Sun's expertise and someone to stand behind your hardware and software (big IT houses), you call Sun and buy their big iron running Solaris. And then, with all the Solaris compat in the linux kernel, it is just as easy to add cheaper boxes doing all sorts of other work using Linux complementing the expesive big iron. Since they now (in the future, maybe...) both have transparent application compatibility, and both kernel's do things in a similar way, then the impetus to buy M$ is greatly reduced. Less IT training, more people who can administer both sytems. You haven't really boosted a competitor, but rather given yourself a new market.

      What confuses me about this strategy is that Sun has never been known as a company that would do what's right for the consumer, as much as it would say what's right for the consumer. For a while they could get away with it with little problem. Maybe now, given M$'s predicament, and that Linux isn't going away, they have to rethink their own borg strategy, even if it wasn't as blatant or heavy handed as the devil...oops, Bills.

      Anyway, as long as they don't have visions of forking or tuning the kernel in a very sun branded way, then this is probably a good thing. If they have visions of steering linux with Scott at the helm, then we may have problems. Hope for the best, so the disappointment can make you stronger.

      • What confuses me about this strategy is that Sun has never been known as a company that would do what's right for the consumer
        That is the necessary paradigm shift. More and more computer systems will be interconnected and the various suppliers had better play nice or the customers will be looking elsewhere.
  • by Anonymous Coward
    You can always tell a company is on it's way down when it announces a major push to embrace Linux. Usually it means its products are not selling and it has to do something *anything* to keep afloat. This is not a good sign for Sun.
    • I guess they can join IBM in the ol' "Spiral of Death" eh?

      Puh-leeze. IBM has wholeheartedly embraced Linux and is stronger than ever.

    • They realize Linux is popular and are building a bridge between their own high end stuff and more low end linux stuff. They are also improving their chances of being able to bail on Solaris and switch to Linux if that need arises in the next five or ten years. They also want to bring things like SunONE to Linux so SunONE will gain share over .NET. Long term, as mentioned in another recent interview, they'd like the OS of a machine to not matter much at all. Instead, your whole datacenter would be managed as some sort of super-cluster with Sparcs and x86s and whatever all running your favorite flavor and all working together and managed centrally.
  • by Geekonomical ( 461622 ) on Tuesday May 14, 2002 @12:03PM (#3518092)
    Guidelines for posting in /.

    1. Scan for news items that has the keyword Linux
    2. Cut and paste a few lines from the story

    Add lines lis (*) is in the right direction for acceptance of Linux.

    Anybody thinking about writing a perl script for this purpose? Lets call it postbot.
  • While it is better than nothing, it sounds like duplication of effort — why keeping improving a proprietary piece of software if eventually GNU/Linux and the Hurd are bound to overcome whatever advantage it currently has?

    OTOH, this could be trying to give Solaris a lease of life while they prepare a real migration plan to GNU software.

  • I thought about possible names to give this porting project/hybrid... Lolasis (sounds too suggestive...) Sinux (sounds like the devil's OS... wait a minute thats M$-Windows, sorry) After my experience trying to install it on an x86 machine... SolarisSUX But wait I digress; what impact will this have on their SPARC chips...
    • This reminds me of a screenshot my coworker took which I keep printed on my office wall. There's a yes/no dialog box asking to reboot Solaris after installing some software on it (cant remember what it was, but still).

      There's also my coworker's hilarious comment beneath it : "WINOLARIS ???"
  • by aralin ( 107264 ) on Tuesday May 14, 2002 @12:05PM (#3518120)
    I'm software engineer, I have both solaris and linux workstations on my desk and to be honest, I prefer to work on my linux and run my database and hard computations on my Ultra-1. Why? Well, Solaris sucks compared to Linux in the ease of use and ability to script your work easy, man pages, utilities, ... and many more aspects. It might be superior OS in matter of stability or effectiveness to use the SPARC platform, but ...

    So they did the logical step. Looked in what is Linux better and try to incorporate these things in Solaris. I say, way to go. But its not to increase Linux's acceptance, really :)

    • Well, Solaris sucks compared to Linux in the ease of use and ability to script your work easy, man pages, utilities, ... and many more aspects.

      What makes you say that?

      You're using CDE, aren't you?

      You're aware that OpenWindows will happily run any window manager, right? And that any utilities you want to install will almost definitely build straight out of the box? And that you can use NetBSD's Zoularis if you want package management?

      At least, you must know about http://www.sunfreeware.com/ [sunfreeware.com], I hope?

      Solaris is actually a very good OS for a workstation. Its X implementation is really fast. Granted, the window manger they ship completely blows, but nobody in their right mind would ever user CDE anyhow.

      Give it a chance before you just presume it's crap because it looks like crap on the surface.
      • by aralin ( 107264 ) on Tuesday May 14, 2002 @12:29PM (#3518315)
        You know, in large IT company you usually do not decide over what you have installed on your workstations. Said that, I'm actually using Gnome on Solaris (looked best for me) and I don't really criticise the window interface. CDE would be fine for me, since all I do is open several shell windows anyway and mozilla in one screen.

        The problems are more in the utilities that are missing some of the more useful switches, man pages that lack behind the linux ones usually, problems with handling symlinks and hardlinks until 2.8 and other minor things that just get on your nerves with time.

        Of course, I can use the GNU utilities, but I cannot write a script using them, because I cannot rely on customer having them installed so I just have to deal with the solaris's crappy ones anyway... do you get the point?

        • On the other hand, Sun's AnswerBook far outstrips all Linux documentation for ease of use and completeness. Linux needs to take a leaf from Sun and Microsoft's books of usability when it comes to providing documentation (both for developers and end users).

    • Well, Solaris sucks compared to Linux in the ease of use and ability to script your work easy, man pages, utilities, ... and many more aspects. It might be superior OS in matter of stability or effectiveness to use the SPARC platform, but ...

      Actually, what I really like about Solaris is that most of the utilities aren't GNU, and thus don't have that asinine "the real documentation is in the info pages" excuse for a manpage that plagues most linux distros (I understand some distros run info2man to correct that ... not sure which ones). What with the utils not being gnu of course makes for some under-featured utils, but I try to not use gnu-isms anyway if I can help it.

      No, the all-in-one-giant-page format of manpages is not ideal for complex apps (it got really ridiculous for perl before they split it up) but it is handy for most utilities.

      NFS that works out of the box with all the wizzy features like failovered mounts is also nice...
  • Me too! (Score:3, Informative)

    by hangdog ( 8755 ) on Tuesday May 14, 2002 @12:06PM (#3518126)
    IBM's AIX 5L [ibm.com] has Linux integration...available now.

    The "L" stands for Linux Affinity.
  • by Jah-Wren Ryel ( 80510 ) on Tuesday May 14, 2002 @12:07PM (#3518137)
    This is at least as much about Linux as it is about competing with IBM and HP. IBM released AIX 5L where the L stands for Linux - they tried to re-implement as much of the linux environment as possible in the AIX kernel and include a bunch of GPL utilities. HP has got a linux porting environment or something like that which is mostly a port of glibc and headers plus utilities to HP-UX 11i. All three vendors have the same goal, to keep their proprietary unix from being completely replaced by linux.
    • Looks like Sun have taken to heart the phrase "If you can't beat 'em, join 'em". Wouldn't using Linux just make their products more competitive which in this era is what all tech companies want to be?
    • Right. What these vendors are doing is commoditizing Linux. The idea is that if you put all of the GNU tools, including glibc, gcc, GNU make, etc., along with some of the stuff typically shipped with a Linux distro, such as gtk, glib, Qt, etc. on a commercial OS, you can then compile most Linux apps (theoretically) without rewriting any code.

      Of course, OTOH, a lot of Linux apps are already written, with the help of the magic of GNU autoconf, so that they can compile on a commercial UNIXes using vendor-supplied ANSI C compilers.

  • perhaps they have found that porting GNOME to solaris was more difficult than it could be and have decided to intigrate features of the GNU/Linux OS into Solaris in the spots that made it tough.
  • Hm... (Score:5, Interesting)

    by Wakko Warner ( 324 ) on Tuesday May 14, 2002 @12:09PM (#3518159) Homepage Journal
    IBM did this with AIX a year or so ago, for 5L. You can (theoretically) compile any Linux application on AIX without any source code modification, and 5L ships with tons of Linux/GNU tools now, a lot of which are installed by default. Linux is suddenly (and rather quickly) becoming more than just a buzzword in the Real World (i.e. not just Internet companies.)

    - A.P.
  • Wrong!! (Score:2, Insightful)

    by rabbits77 ( 453747 )
    This is like saying that MacOSX has opened doors for *BSD in the graphics art niche. Technically, it has but most of the people using it don't care and see it as 'Mac'. Same thing here. 90% of their customers will just upgrade to the new rev of Solaris and the fact that it has something to do with Linux will be almost completely ignored. No more than 1 bullet in the sales pamphlet.
  • As a first step, they might make their system calls and programs have the same interface (parameters) than Linux.
    • Re:Solaris and Linux (Score:3, Interesting)

      by Anonymous Coward
      Given that Solaris has versioning in its dynamic loader I'd be willing to bet they will add Linux API's to Solaris.

      Solaris already has most of BSD's API's if you know where to look so its not too much of a leap to add Linux versions as well.

      Solaris is well ahead of BSD and Linux in most kernel technologys, its main drawback is ooey, chooey GUI stuff and thats probably what they are actually going to try and make "compatable".

      Look at most major open source packages older than 3 years and you'll see Solaris already is supported, its only in the more recent packages that Solaris isn't fully supported and usually because the people developing the packages don't know where to look in Solaris' API's; i.e. which header files and librarys to use.
  • sunux? (Score:3, Funny)

    by tps12 ( 105590 ) on Tuesday May 14, 2002 @12:16PM (#3518201) Homepage Journal
    I have already had the idea to combine Linux and BSD to create LSD. This sounds similar.

    Sunux? Solarux? Linaris?
  • by Gerdts ( 125105 ) on Tuesday May 14, 2002 @12:16PM (#3518207)
    This looks like a step in the right direction for Linux acceptance in the professional server market.
    This is not what Sun is intending to do. They realize that there are a lot more people out there that are writing code for Linux and making it a desirable OS to use than there are doing the same for Solaris. By making it easier to port from Linux to Solaris, Sun is trying to make it easier for developers to keep Solaris as a top tier platform.

    For instance, if you have installed OpenSSH on Solaris, you will have been forced to look into the various methods for getting /dev/urandom or a suitable replacement. After I brought this issue up and reminded Sun that they were trying to get to a Linux-compatible API, they backported their Solaris 9 /dev/urandom to Solaris 8 with patch 112438-01 [sun.com]. Imagine my shock that Sun actually implemented one of my RFE's.

  • by I'm Spartacus! ( 238085 ) on Tuesday May 14, 2002 @12:17PM (#3518215)
    Here's the configuration for our largest server:

    HOSTNAME: grande, OS: SOLARIS 5.8, MACHINE TYPE: E6500 , USER: Server
    MEMORY: 28GB, SWAP: 9GB, PROCESSORS: 28 400MHZ, DISK: Fibre Channel Raid 136GB

    Linux can't come close to this kind of setup, and I doubt it will anywhere in the near future. Now admittedly, Linux is hurting Sun in a big way. Sun hardware is damn expensive. But we need that kind of hardware here in our shop, and Linux simply won't cut it.

    Sun is doing this because Linux is hurting them on low end hardware, not because Linux is in any way better than Solaris for anything other than skinning your desktop.
    • But it's in Sun's best interest to pave a smooth superhighway upgrade path from Linux to Solaris for users that grow beyond their x86 hardware

      Oh yeah?? Well look at *my* setup:

      HOSTNAME: mofo, OS: SOLARIS 9.2, MACHINE TYPE: Z69000 , USER: bofh
      MEMORY: 64326GB, SWAP: 52376GB, PROCESSORS: 5800 2.5GHZ, DISK: Fibre Channel Raid 593271GB


      • HOSTNAME: mofo, OS: SOLARIS 9.2, MACHINE TYPE: Z69000 , USER: bofh
        MEMORY: 64326GB, SWAP: 52376GB, PROCESSORS: 5800 2.5GHZ, DISK: Fibre Channel Raid 593271GB

        Yeah we've had one of these for a few years now. They make great routers/firewalls to keep our REAL machines free for the serious work...

    • Here's the configuration for our largest server:

      HOSTNAME: grande, OS: SOLARIS 5.8, MACHINE TYPE: E6500 , USER: Server

      MEMORY: 28GB, SWAP: 9GB, PROCESSORS: 28 400MHZ, DISK: Fibre Channel Raid 136GB


      Nothing, in comparison to 2 E10K's I've played with last week. And you know what? That is nothing to GS320 with 256GB RAM and 32 CPUs, which had Linux running on it (check linux-kernel mailing lists).

      Bottom line is - Linux runs on big hardware. It's just that you never tried it.
    • OK, you want a Solaris-vs-Linux pissing contest. Fine. All I ask is that you put your money where your mouth is.

      Download my DNS server, MaraDNS. Compile and run it on Solaris. Fins the problems that MaraDNS has on Solaris and fix them. I have ported or seen my application ported to various OSes, including FreeBSD (it's in their ports tree), MacOS/X, and Windows (with cygwin). The only port that was less than trivial was the Solaris port.

      It took me about a day to get it to compile; then it wouldn't run at all. Running truss revealed that any network application needs /dev/udp and /dev/tcp in the chroot() environment; any multithreaded application needs /dev/zero in the chroot() jail. Things no other OS I ported MaraDNS to needed.

      Even after getting it to run, it would crash when doing even a very simple strees test that the Linux (and MacOS, and FreeBSD, and Windows) version can run without problem.

      This was on Solaris 7 (x86).

      If you want to impress me with how great Solaris is, I would deeply appreciate any help you could provide making this application as stable on Solaris as it is on all of the free *NIXes.

      Then, and only then, would I feel that Solaris was a useable OS with a helpful community; right now I see Solaris as a buggy OS with an arrogant community that I don't want to be around.

      Thank you.

      - Sam

    • I guess you missed the slashdot story a while back which claimed someone was able to compile the Linux kernel on some ungodly 32-way numa machine in under 8 seconds! That's right, 32-way! Linux is making _huge_ progress in the 2.5 developement series, such as the new O(1) scheduler, NUMA support, Block I/O, you can check out the 2.5 status page [kernelnewbies.org] to see it all. Perhaps 2.4 can't scale like solaris can, but 2.5 will kick ass once it's ready. In the mean time, alot of the important patches are being backported to 2.4 (such as the new scheduler), I recommend the JAM [unizar.es] patchset.
    • How about this configuration:

      OS: Linux, MEMORY: 1.8TB, PROCESSORS: 1400 Itanium, DISK: 170TB

      Wondering if you'd seen this before? Here's your friendly reminder [slashdot.org]...

    • "Inferior" at what? I work in a mixed environment myself: Linux, *BSD, Solaris (6,7,8), and (*cough*) Win2K. While I agree 100% that Solaris scales to multiple processors at levels that Linux and *BSD can't touch, for the desktop I'd have to say Solaris is inferior - which is the point you're making.

      Perhaps, by incorporating some of the more 'desktop friendly' aspects of Linux into Solaris, Solaris will become less of a sucky proposition on the desktop - which is where most of the world's computer users live.

      Murphy knows trying to put Linux on an E65k would be nuts. Sun's Heavy Iron runs best on Solaris and usually isn't saddled with desktop apps. But for the Light Iron (eg: old Ultra5's, modern SunBlade 100's) - which are cheap by Workstation standards - a lot of the more friendly Linux features make sense.

    • At work we use our machiens for some pretty heavy math computations... and we're not really looking into Solaris anymore. The fact is, that e6500 costs a SHITLOAD, and for similarly priced low-mid range hardware, Linux wins hands down.

      Actually, I'm quite curious -- what do you guys need that monster machine for. You don't have to give away everything, but I can't really imagine too many uses for something that heavy. (esp. that couldn't be done with a more distributed architecture)

  • by Anonymous Coward
    What about the x86 version of Solaris? Are there plans to implement this on that platform or is x86 Solaris dead? Also another point of interest, sence Microsoft has a version of IE for Solaris how long do you think it will be until IE is able to be run alone on linux?
  • by mikeee ( 137160 )
    I'd like to see User-Mode Linux ported to Solaris. Dozens (hundreds?) of Linux boxes running on a single Solaris machine... yum.

    Is this actually feasible, or am I on crack?
    • Re:UML (Score:3, Informative)

      by ianezz ( 31449 )
      I'd like to see User-Mode Linux ported to Solaris

      In theory it should be feasible (Wine proves it can be done even for two completely different systems).

      In practice I wonder how much overhead you are going to pay (I keep hearing that system calls on Solaris are much more expensive - and consider that each system call in UML in turn would be implemented as several system calls to the hosting system).

      GUI shouldn't be a problem: interactive applicatione are usually 99% idle anyways, and using them should be as simple as an "export DISPLAY=..."

      OTOH, I/O bound processes probably would be penalized too much, and it would be a good idea to execute then directly on the hosting system.

      In the end, if the ability to be root in your very own "partition" is worth the (hypotetic) additional overhead, I'd say "why not?" Of course, some numbers are needed here...

  • This is the sort of thing where the GNU/Linux distinction becomes significant. It isn't very clear about precisely what is being copied/encorporated, but if they want to easily run software from 'Linux', they will need pieces of both.

    For anything remotely shellish, they will likely need the GNU file-utils and text-utils. This would, IMHO, greatly improve Solaris anyway. They already include bash, gcc, and emacs (though they do ship their own shell and compiler as default) and are already planning to include GNOME.

    In short, Solaris already includes massive parts of GNU.

    Now Linux is a somewhat different issue. Duplicating kernel APIs is pretty new (by Sun of Linux, that is). It shouldn't be that big a deal, though -- there is still POSIX underlying everything.
  • by bachelor3 ( 68410 ) on Tuesday May 14, 2002 @12:32PM (#3518331)
    A quote from Mehra, who has the "Linux advocacy role" at Sun:

    "Linux is where Solaris was five or 10 years ago."

    You make your own cynical comment : )

    • "Linux is where Solaris was five or 10 years ago."

      You make your own cynical comment : )

      Okay, I will. How about, "That's pretty accurate, and most anyone who uses both Linux and Solaris on a daily basis will agree."

      Possibly not the cynical comment you were looking for :-) but the design and stability of Linux has a long way to go to catch up with Solaris 8. As far as cool'n'nifty user features go, I can just compile the occasional GNU utilities on my Solaris box.

      • It depends a lot on what exactly you're doing. Solaris is better for support for 'big iron' things like RAID, there is nothing like Veritas's VXVA on Linux, or for logical paritioning. On the other hand, Solaris's /proc implementation isn't as good as Linux's, which can make a big difference in day to day usage of Linux.
        • The Linux software raid and LVM support are actually fairly comparable to Sun's free Solaris equivilant (Disk Suite). This changes with the new version now called Solaris Volume Manager howerver, which introduces "soft partitions" which makes it more like VxVM. This should be bundled with Solaris 9.

          VxVM and VxFS have been ported to Linux, though who knows when it will ever be publicly released. RedHat has been bankrolling the port and the release should be available for their Advanced Server product. I have an early beta version that only compiles against a 2.4.0-pre kernel but it works. I think the many changes that have occured as 2.4 has only recently stabilized have probably held things up. Veritas is doing a lot of kernel work and I'm unsure how widely it would be accepted by the community if/when it's released since it won't be entirely open or cheap.

          I think the real areas where Solaris far outshines Linux include scalability beyond 8-procs and things like hotswap CPU support. But all these areas are being actively addressed. The new scheduler in 2.5 should go far towards addressing scalability and pretty much everything else is being actively worked on. Given the fast pace of Linux development, the real gap is probably only a year or two. If the big iron issues don't affect you and you don't need anything >4 procs, I don't think there even is a meaningful gap. Just massive $$$ savings and easier administration (at least w/Debian).
          • If you've got a problem which requires it, then it's definatly worth the money for Veritas, which is why it's still being sold even with the free package available. I'd imagine this would remain the same, even for Linux. However, there are a lot of products in the product line, and until they're all available, then I'd not consider Linux a possible replacement. HSM & FlashSnap are part of the reasons why I want to use Veritas.
        • there is nothing like Veritas's VXVA on Linux, or for logical paritioning.

          Ahem. I assume you mean VxVM (which is what does the partitioning). And since you ask, there are three things that do that on Linux: LVM [sistina.com], EVMS [ibm.com], and yes, genuine Veritas VxVM [veritas.com] (and also VxFS thrown in, to boot). Solaris may be better at some things than Linux, but the number of things that fall in that category is shrinking rapidly. Volume management no longer qualifies.

    • "Linux is where Solaris was five or 10 years ago."

      You make your own cynical comment : )


      Thank you, I will:
      Ya mean Linux is the runaway growth leader in the production server OS market? OK, I'll buy that.

      :-)
      • No, in many ways she is right. The jewel in the crown of Sparc/Solaris is the RAS features, which they justifiably use as a stick to beat IBM/HP with.

        Solaris/Sparc allows automatic system recovery - if your machine crashes with a hardware fault, it will find the fault at POST time, and boot the kernel around the faulty hardware. Not such a good thing when thieves broke into XXXX organisation, nicked a load of system boards leaving one in each machine, and the machines recovered and restarted the applications running on what hardware was left.

        Solaris/Sparc allows you to add and remove CPU/memory boards to and from a *running* domain.

        And Solaris/Sparc has a SMP kernel which will scale linearly to over 100 processors - this isn't just a particular design choice about the kernel architecture, it's a lot of choices about the hardware, firmware and kernel details.

        So while the Linux kernel has proved itself massively capable for horizontally scalable systems, the Solaris kernel has been designed for vertical scaling. There's actually no need for the Linux kernel to head in this direction - it's a good solution where you can have multiple boxen scaled out, or where you are using a HA/clustering solution, but for large installations with (for example) massive database instances and requiring the RAS features which Solaris/Sparc offers, the Linux kernel doesn't fit.

        Dunstanb
  • Note, it's their Linux guying talking. When McNeally gets up and says we're merging Solaris and Linux, then you can get up and pay attention. This is just another step in Sun's long, slow embrace of Linux on the x86 platform. So long as they make most of their money from SPARC boxes running Solaris, I don't expect you'll see Linux at their core. Running Linux programs in a compatibility mode, ala Caldera's OpenUnix, yes; running a Linux kernel on SPARC, as the recommended course, no.

    Steven
  • This looks like a step in the right direction for Linux acceptance in the professional server market.

    Only somebody with zero Unix, and hardly any linux would say such a stament. The fact is that SUN doesn't give a damn about Linux, it jsut wants the exposer... Suns strategy is to Maintain Solaris 9 for the server environment, and deligate a Linux kernel for the desktop space. And don't let the word "Linux" fool you either... Linux is a kernel program, not an Operating Environment. So yes, Sun plans to sell Solaris 9 in the server space, and sell solaris with a linux kernel (possibly) for the Intel x86 systems on the network. It is possible SUn may make a distro of Sol9+linux for Sparc too, but who really cares. Most of the stuff that would make me want to use Linux on a Sparc box is now a default feature in Sol9.

    The big mess Sun got into when they anounced they were dropping Solaris 9 for x86 ARCH wasn't such a suprise to me, considerign they have for a year now been say they are going to develop their own Linux distro to handle that segment of users. I wish people would wake up and pay attention. This is such old news!
  • by NorthDude ( 560769 ) on Tuesday May 14, 2002 @12:57PM (#3518538)
    Generally however we've found that the cost of open sourcing code for a proprietary product is non-trivial. I know it seems counter-intuitive but consider this: the reality is you can't just toss code over the fence.

    Open-Sourcing Solaris is non-trivial, she explain it in her answers. But working on already open-sourced coded is not. It benefit both Sun AND the Linux community because Sun's change will get back to the community and it benefit sun because they have a very solid base to work on.


    Sun's position on Linux has long been friendly, since we see it as a commodity unix variant which has been very successful at growing the community of Unix users.

    I wonder how much "has long been" really is, but it's not the point. I found it is rather honest on their part to say it that way. The first 8nix variant I saw was Mandrake 6.0 or somehing like that. I felt in love with it and since I had the chance to deploy aaplication on Solaris a couple of time. So the comment makes sense, Linux has a lot of visibility and it happens sometimes that it is what brings users to the realm of Unixes. So, even from a marketing point of view, it all make sense to adopt it. It gives them free publicity because of their implication with Linux. And afterward, it benefits them because they can either sell more Solaris or just more server, even with Linux on it instead of Solaris, they make te buck with thehardware.


    All that being said, they had to previously (well, they have to) "support" both Linux AND Solaris and port appliation to both platform. By trying to standardise both, they keep the previous;y stated benefits, and they do a cut in the devellopment budget.

    And in the end, it benefits us to. That's is the way I would like all business to work. Make your own business, cleanly, and work WITH the community. It can only do good, both for the business and to the consumer/user/geek/etc etc...
  • by jellomizer ( 103300 ) on Tuesday May 14, 2002 @01:12PM (#3518678)
    This move is good for Linux, is good for Sun, its good for IBM it good for HP and it is Bad for Microsoft. When you try to talk to middle managment about using Unix systems they come up with the excuse that there is little software for that platform. And Comerical Developers will more likely program for MS Stuff becuase that is where the market share is, ms is reported to have between 30-45% Marketshare), Now Add the big UNIX guys Combined make up about 50% of the market share all start playing nice with each other, Sure they are compeating against each other but they try to make a better product the the other guy, And agreeing on a Linux format API. Why Linux because Linux is not owned by any company so you are not giving one company a head up on the other. With a simular API style it is easier for Comerical Programmer to make programs for the different platforms so say I made FooBar Server program on my Linux box that the source code can port super easy to a Solaris, HPUX, AIX box it a good thing because my FooBar Server can be accessed on 50% of the servers.

    Consumers get the benefit because there are more and cheaper programs available for their Platform. And they can choos the type of *X platform they want to use.

    The UNIX companies get extra insurence that there is a chance that they can get repeat business form there Customers. And have the advantage of more software for their platform.

    Smaller Developers and Support personal get the advantage of easy comunication between the different Unix systems.

    But it will hurt the following people.
    Microsoft. Becuase they are being "more" seporated from the curent standards. And being shunned my more third party developers.

    Windows only programers. But it is there fault for not following the real standards. And opening there mind into more cross platform development.
  • This looks like a step in the right direction for Linux acceptance in the professional server market.

    Huh? Every couple of weeks, there is another story in the news of some big company dumping their Suns for x86 servers running Linux. Where did you get the idea that Linux is not already accepted in the "professional" server market?

  • Gee, adding Linux compatibility to a proprietary Unix. Isn't that exactly what Big Blue [ibm.com] has done with AIX 5L [ibm.com]?
  • by timbrown ( 578202 ) <slashdot@machine.org.uk> on Tuesday May 14, 2002 @02:01PM (#3519041) Homepage
    Sorry, but I think you Linux evangalists are plain wrong. I work for a company that supplies managed services to blue chip firms and the demand is still there for Sun boxes, due to the quality of the OS and hardware they ship. Linux may have the some features of Solaris, but it doesn't *yet* have the track record and enterprise level support that our clients require. The worrying thing is that whilst for smaller shops Linux is being used for low end web servers etc, for our customers they'd rather use W2K in the cases where they can't justify the cost of a Sun box :
    A few things I like from Solaris that Linux doesn't really have yet... Scalability, I know its not an issue for most of you guys, but Suns 106-way boxes are really quite neat. Technologies such as JumpStart, which make rolling out a new web cluster a breeze. Stable IPv6/IPSEC support. Comprehensive support, from *one* source. A top class architecture to run the damn thing on.
    I like Linux, don't get me wrong, I personally have 2 Debs boxes and manage a Slack box in Slovenia, but I also have a FreeBSD box, Sparc running Solaris 8 and a HPUX powered PA-RISC machine.
    My attitude is that if it has /bin/ls I like it, but of all the UNIX like platforms I've worked on Solaris is my favourite.
  • ...and start telling people to say GNU/Solaris.

Help me, I'm a prisoner in a Fortune cookie file!

Working...