Stories
Slash Boxes
Comments

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

IBM Open Sources Supercomputer Code

Posted by timothy on Wed Aug 06, 2008 02:48 PM
from the buttered-up-penguin-is-delicious dept.
eldavojohn writes "IBM has announced at the LinuxWorld conference that they are now hosting all their supercomputing stack software as open source from the University of Illinois. From the article: 'The software will initially support Red Hat Enterprise Linux 5.2 and IBM Power6 processors. IBM is planning to add support for Power 575 supercomputing servers and IBM x86 platforms such as System x 3450 servers, BladeCenter servers and System x iDataPlex servers. The stack includes several distinct software tools that have been tested and integrated by IBM. These include the Extreme Cluster Administration Toolkit (xCAT), originally developed for large clusters based on Intel's commodity x86 architecture but now modified for clusters based on IBM's own Power architecture. xCAT is used in the National Nuclear Security Administration's Roadrunner Project at Los Alamos National Laboratory in New Mexico — a hybrid cluster currently ranked by the official Top 500 list as the world's most powerful supercomputer.' For several years, Linux has been a strong tool for supercomputing."
+ -
story

Related Stories

[+] Linux: Linux Claims 4 of the Top 5 Supercomputer Spots 253 comments
Anonymous Coward writes to tell us that the November 2005 list of supercomputers has been published. Certainly something to note is that four of the top five use linux. Relatedly Multiflow writes "CNET is reporting that the number of supercomputers on the Top500 list which use Intel Itanium 2 microprocessors has fallen by almost 50% in the past year. While new higher performance Itanium chips are in the pipeline, the article reports that 64 bit Xeons and Opterons have increased their representation on Top500."
This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
 Full
 Abbreviated
 Hidden
More
Loading... please wait.
  • Sweet (Score:5, Funny)

    by 0100010001010011 (652467) on Wednesday August 06 2008, @02:49PM (#24501745)

    Now I have something to run on that spare Power4 I have laying around in the basement.

    • Re: (Score:2, Interesting)

      Seriously, though: would the Power6 have been the successor to the chip family we call Apple's G5? (You know, if Apple had continued using PowerPC chips?) I'm researching this using Wikipedia but haven't yet discovered the answer...any Slashheads know?
      • Re:Sweet (Score:4, Informative)

        by pleappleappleap (1182301) on Wednesday August 06 2008, @03:17PM (#24502189) Homepage

        No. The PowerPC architecture is mostly a subset of POWER. The POWER processors have all of the instructions of PowerPC, plus more. And they're built a little more robustly, since they're designed for the enterprise server market.

      • Re:Sweet (Score:5, Insightful)

        by EvilRyry (1025309) on Wednesday August 06 2008, @03:17PM (#24502191) Journal

        Something power6 derived anyways. Apple always wanted their chips with the Alitvec instructions which weren't part of any of the other power series. They also didn't want to pay a whole lot for these custom chips which they order in relatively small quantity. Its little wonder IBM didn't rush to get them new CPUs, they're probably happy Apple is just leaving them alone.

        • IBM's later chips have AltiVec, even though they have nothing to do with Apple anymore.

        • Re: (Score:3, Insightful)

          Really? Small quantities? IBM is really selling so many supercomputers that they need to produce more than a million of these Power chips every year?

          More likely, the requirements that Apple wanted/needed for it's PowerPC chips were different enough from where IBM wanted for the Power chip line, and IBM/Apple couldn't come to a financial agreement to produce the PowerPC chips that Apple needed [as in, the combination of cost per chip/capabilities of chip/when chip would be available that Apple wanted and w

          • Re: (Score:3, Interesting)

            If each supercomputer contains a hundred thousand CPUs, they only need to sell TEN supercomputers.

            If each mainframe contains a hundred CPUs, they only need to sell ten thousand mainframes.

            See, those are quantities that help make sense of this. A Blue Gene/P installation can use up to nearly 900,000 processors alone.

            So yes, IBM probably does ship more CPUs than Apple does. IBM doesn't just fabricate and sell Power chips either, so I'd say there's probably a pretty wide margin.

            • Keep thinking that when Apple surpasses 10 Million Intel packaged systems or greater by mid 2009, year over year.
              • Re: (Score:3, Interesting)

                There's an IBM chip in every one of the three major consoles, which have sold around 50 million units in 2 years.

                And that's a small part of their business. IBM sells a lot more chips, period, than probably anyone other than Intel or maybe there's an ARM manufacturer that does more business. But IBM also fabs ARM CPUs, so there you go.

    • I don't know about you, but I actually do have extra POWER hardware in my basement. And my living room. And my dining room.
    • I hear Vista runs decent on this hardware.
  • Great (Score:4, Funny)

    by thammoud (193905) on Wednesday August 06 2008, @02:50PM (#24501765)

    Just when there no longer any COBOL programmers around.

    • COBOL for supercomputers? I don't think so. More FORTRAN. COBOL is used for business code. These machines are primarily used for modeling.

      • These machines are primarily used for modeling.

        You mean like on a runway across the top of the machines? You may have just thought of a way to re-ignite interest in supercomputing.
      • COBOL for supercomputers? I don't think so. More FORTRAN. COBOL is used for business code. These machines are primarily used for modeling.

        Yeah, I think of a single good reason to run business code on an International Business Machine! </sarcasm>

      • in the 70s and 80s COBOL compilers were available from Cray and CDC to run on their supercomputers.

  • We don't host xCAT (Score:4, Informative)

    by dlapine (131282) <dlapine@ncsa. u i uc.edu> on Wednesday August 06 2008, @02:53PM (#24501815) Homepage

    Try here [sourceforge.net]instead. And yes, xCAT kicks butt if you want to run a linux cluster. More so, now that it's open source.

    • Re: (Score:3, Informative)

      by Anonymous Coward

      Uh, yes we do.

      ftp://linuxpatch.ncsa.uiuc.edu/OpenHPC/

    • I sit corrected. IBM has allowed NCSA to host a special version of xCAT for RHEL 5.2 on PPC. You can find it in here [uiuc.edu].

  • So guess we won't have to imagine a beowulf cluster of this. Phew, meme crisis averted.

  • the real goal (Score:4, Insightful)

    by xzvf (924443) on Wednesday August 06 2008, @03:17PM (#24502193)
    I work for IBM, but this is speculation. The vast majority of money generated and earned on large Linux clusters came from selling hardware and services. This can only help generating that business.
    • by mikael (484) on Wednesday August 06 2008, @08:37PM (#24505363)

      I heard this story about Cray. If you bought a supercomputer and support contract from them, they would throw in the free construction of an office block to run the computer in.

  • Just in time, I really needed a supercomputer. Let's play Global Thermonuclear War
  • by contrapunctus (907549) on Wednesday August 06 2008, @03:20PM (#24502235)
    They open source stuff and they patent ridiculous stuff. Am I supposed to like them or not?
    • If you don't want them to have to patent "ridiculous stuff" ensure that the business environment changes so that they don't need to.

          • Sorry, but your comment was similar to:
            If you don't want people around the world to starve, ensure that they have enough food.

            This is not specific enough to be meaningful.

              • But see, I wasn't specific enough to say "give them food" I said "ensure that they have enough food" which could mean teaching or anything I want.

                My whole point was that making vague statements is meaningless. They don't mean anything. They could mean anything.

    • I tend to think they use open source software as a way to sneak high priced consultants in to sell high priced software and support, but I might be being paranoid.
  • Since /. was adding pictures, Farking them is only a natural extension of expression. Supercomputer Dialysis Machine [photobucket.com]
  • patents? (Score:3, Funny)

    by Anonymous Coward on Wednesday August 06 2008, @03:30PM (#24502395)

    will it be offered in paper or plastic ?

  • I mean, it has a SourceForge page [sourceforge.net] whose mailing list archives go back to 2001, fer cryin' out loud.

    Now some of the "OpenHPC" stuff appears to be new, but not all of it appears to originate from IBM. For instance, part of it appears to be a repackaging of the SLURM [llnl.gov] batch system from LLNL. The one thing that looks like a genuine contribution from IBM is the "Advance Toolchain" stuff, but even that appears to draw heavily from existing open source code bases like valgrind.

  • by shlompo (1338043) on Wednesday August 06 2008, @04:28PM (#24503229)
    I used to be in charge of administrating the lab cluster at the MOSIX project (http://www.mosix.org). The tools we used back then, where series of scripts, that performed all possible configurations you'll ever need... we called it CLIP (CLuster Installtion Package). My two years experience taught me two things:
    1. It's sometimes easier to script your way through, instead of adapting existing administration tools. You'll just have a peek first, of course...
    2. But when you must, you'll encounter a modification you'd want very quickly.
    So my advice would be only accept open source administration systems. As i'm sure others have reached the same conclusions i had, This is actually a win-win move by IBM, and i'm sure they'll get more users, and more income following.
  • by unity100 (970058) on Wednesday August 06 2008, @04:34PM (#24503301) Homepage Journal
    did in latest oscon. what do you see ? rock solid commitment compared to empty pr. you know which of them pertains to which company ...
      • Being a supercomputer, I would imagine it could play Crysis comfortably in a pure virtual machine, emulating a graphics card as well.

        • Potentially, but would programming responsiveness [gamasutra.com] not be an issue in a system just designed to digest data, the IPS might be there but there could be more to it then that.
        • But can I play Quake on it?

          Actually, you can. A couple of games [ibm.com] are available on the IBM AIX software FTP site. You have your choice of Quake or Quake 2, even. Since POWER6 still runs binaries from the previous generations of software, go give it a try. Of course, you'd probably need to launch one instance of Quake per thread to really stress the system, since I'm pretty sure Quake's not multithreaded. ;-)

          ~ Mike

    • That would be exceedingly stupid. Why not just make a larger cluster over the high-speed clustering medium instead of throwing unnecessary Beowulf overhead into the process.

    • Better yet, imagine an obligatory Beowulf cluster of spell-checkers.
    • by unity100 (970058) on Wednesday August 06 2008, @04:35PM (#24503321) Homepage Journal

      IBM has been supporting the Open source community but has the community returned the favor?

      • Eclipse. Every summer they do a multi project release on the same day. This year the release was called Ganymede [eclipse.org]. The release announcement follows (emphasis added):

        The Eclipse Foundation and the entire Eclipse community are pleased to announce the availability of the Ganymede Release, the annual release train developed by the Eclipse community. The Ganymede Release is a coordinated release of 23 different Eclipse project teams that represents over 18 million lines of code.

        IBM didn't write all that code ;)

      • There's a difference in OFFERING source code for something, and making it ACTUALLY open source. Slashdot mixes these two all the time.
      • Oh to answer you question: Yes, the community can return the favor by using Java. The reason for IBM "giving away" Eclipse was to get programmers to accept Java.
      • By supporting and contributing to Open Source, they become part of that community, so yes they have returned the favor to at least them selves.

        The community are not just the individuals. It is everybody and everything that contributes. They have the ability to use anything from Open Source. If they desire to do so or not is up to them, not to the community.

        So the question is not if the community has given anything, but if IBM has taken anything and I would say yes. They are using code that is written by oth

        • No, they don't.

          They sell support and services around the distribution. Even packaging the product up onto a CD with some printed manuals can be considered a service.