Slashdot Mirror

← Back to Stories (view on slashdot.org)

Distributed Climate Prediction For Linux: Pending

Posted by timothy on from the clouds-beat-aliens dept.

saintp writes "Inspired by earlier discussions about distributed computing initiatives, I emailed the good folks at Climateprediction.net to inquire about a linux client. And, the good news: There will be one. No update yet to their system requirements, but here's what they told me: 'There is certainly a Linux version planned (in fact, already existing), but we are a small team and can only set up the infrastructure to support distributed experiments under one O/S at a time. Running climate models is substantially more ambitious in terms of machine requirements, data generated, security headaches etc than any other distributed computing project we know of, so developing a generic O/S independent client is simply not feasible. As soon as the Windows version is safely launched, the Linux version will be our next priority.' Hopefully, we'll see it sometime in November or December."

11 of 18 comments (clear)

  1. Why? by Guspaz · · Score: 2

    Forgive me if I don't know much about it, but why should it be so hard to make it platform independant? I would think most of the proccessing to be done would be just a recompile away from another platform, and sockets on *nix are virtually identical to on the PC. So what's the big thing that's making this so difficult for them?

    1. Re:Why? by MindStalker · · Score: 3, Insightful

      Basically it boiles down to the fact that mallicious users sending false data would completly destroy this project. In Seti, if some mallicious user sends a I have found ET packet, it can easily be checked and confermed (which I believe is automated even) same with all the other distributed systems so far, a user sending back important info is rare enough that it can be checked. While in weather predictions each user will probably be solving a piece of a puzzle so each users will be sending back significant amounts of info not just occasiall rare users. So each user can't be checked. So the checking mechanism has to be done client side, a virtual impossibility if you ask me.

    2. Re:Why? by dacarr · · Score: 2
      Because it's not running Java.

      The thing is when you're dealing with executables, and I'm not sure why this is, but for some reason your windows code that compiled just fine under GCC for Windows will need tweaks to move over to (say) OS/2's or *nix's respective GCC ports, or as is the case with OS/2, require an added compatibility layer (the EMX runtime libs - what few OS/2 geeks there are will know these very well). It will likely be the case that you will have to do some significant code massaging in order to get it to work.

      Mind you, it's been a time since I did cross-compiling, so AMMV these days.

      --
      This sig no verb.
  2. Distributed computing... by His+Excellency · · Score: 2, Insightful
    ...irks me a little. When I give away my CPU cycles what do I get in return? I get a product that I helped bring to life, but I also have to pay for it.

    Weather reports are not bad though, I just pay for them by viewing annoying advertizing. The projects that really annoy me are the ones used to design drugs to fight cancer or some other like disease, because the information from these projects will be used by pharmaceutical companies to sell me drugs!

    If we are going to use ditributed computing to solve problems, the solutions should really be free or discounted for the general public, because they donated their computers cycles, and shouldn't be forced to "pay" twice.

    1. Re:Distributed computing... by Kj0n · · Score: 3, Funny

      ...irks me a little. When I give away my CPU cycles what do I get in return?

      Well, one thing you get in return is some extra heat and the certainty that you just helped change the climate by using more energy.

    2. Re:Distributed computing... by kippy · · Score: 2, Insightful
      The projects that really annoy me are the ones used to design drugs to fight cancer or some other like disease, because the information from these projects will be used by pharmaceutical companies to sell me drugs!

      as you can tell from my sig, I disagree with your stance.

      I'm participating in United Devices to help find a cure for cancer.

      Am I paying for it? Sure, but the electricity that I'm using is probably less costly then when I forget to shut off the bathroom light when I leave for work. It's trivial, it's pennies, get over the cost issue.

      Will I pay for the cure once it's found? Could be. UD is giving their results to Oxford if I'm not mistaken. From there it might get patented and sold by big industry. So what? What medicine do you currently get for free? except for laughter, none. if you participate in such a project, do you expect the cure to be free for you or do you just want a 0.000000000012% discount for the 0.000000000012% work that you put forth to help find the cure.

      Also, if cancer is curable, the drug will probably be found eventually. I have a friend with skin cancer and my grandfather died of cancer. I smoked so that puts me at a higher risk. I'd rather the cure be found sooner than later.

      Just because something has to be paid for doesn't mean it's a bad thing or that the little guy is getting screwed.

      I'm probably going to get a bunch of responses to this from the "knowledge is free" crowd. Fuck that. If I spend $6 for my energy bill and have to pay $1000 for some magic cancer pill down the road, I'll have a magic cancer pill when all is said and done. If distributed drug simulation isn't done, there's a chance that I or a loved one might die at 80 lbs with no hair. I'd like to know that I did all I could to prevent that situation.

      in any case, it sure beats spinning my wheels looking for ET.

      --
      Blaze a trail to the New World
  3. Whither Java? by Matey-O · · Score: 2

    Why aren't these folks using Java? Remember write once, run everywhere? It's got an authentication mechanism, and the whole 'you can converse with the originating server' schtick seems to make it ideal.

    --
    "Draco dormiens nunquam titillandus."
    1. Re:Whither Java? by Omegalomaniac · · Score: 2, Insightful

      I'm guessing they are not using java because distributed computing is very much about speed. A few distributed computing clients have hand written (or hand optimized) assembly code running the bulk of the computation. Java would be a serios step backwords.

    2. Re:Whither Java? by ianezz · · Score: 2
      Why aren't these folks using Java?

      Uhm, perhaps there could be subtle issues on how various JVM implementations handle floating point math, despite of the Java specifications? I'd say that when you're doing distributed computing, you don't want to end comparing apples and oranges.

      A quick search on Google comes up with something interesting on the subject (which of course may be true or not).

  4. Business model? by Bluedove · · Score: 3, Interesting

    How's this for a preliminary stab at a distributed computing organization business model:

    Each computing unit (whatever size) is allocated a price based on demand. You can use the units at the given price, or you can generate the computing units (have other peoples jobs run on your machine) at $0.50 on the dollar. Your account is tracked/stored with the organization.

    The benefits are as follows:

    -Your spare computing units are automatically contracted out at a competitive price. You can (arguably) make money doing it with all those spare computers laying about.

    -You can stockpile computing units to be used later in your research. (Make your computer work for you even while in the reading phase).

    -If you have a project where you need units, but don't have time to contribute, you can simply buy the distributed processing time.

    -You could enhance your effective computing time by contributing at peak times (when the price is highest) and processing at off peak times (when the price is lowest). You will effectively have improved your computer.

    --
    Demonstrant's Open Source Tools
  5. A sep. DC idea by billmaly · · Score: 2

    Does anyone know if running DC to examine NEAR objects is feasible/practical/needed? This would be to assist researchers with tracking of "doomsday" asteroids that have a chance of hitting Earth. I know that there are not enough people actively watching the sky right now, and we are not currently prepared to engage an asteroid. Still, if we knew one was coming, steps maybe could be taken. Or we could argue about it.

    Bottom line, would DC to track and predict paths of NEAR's be of any practical use, or not?