Slashdot Mirror
Donate Spare Cycles for Climate Prediction
gampid writes "The BBC has a story about the Casino-21 project which is running a SETI@home type program for climate prediction. " I'm a booster of Distributed.net, but this looks pretty cool as well. I dunno global warming just gives me the creeps anyway and I'd like to know if my house will be underwater.
I get calls to donate "spare" money to this and that charity during dinnner hours on my private phone.
My "spare" organs are going to be donated in the case of an untimely death.
People keep trying to bum "spare" cigarettes and gum and change from me.
My "spare" clothes are being worn by freinds and family.
My dog and cat hover over me every meal as if to say "buddy can you spare some of your food"
A crazy chick wants my "spare time" so she can whine to me about her Jerry Springer Guest of a boyfriend.
Now Seti and/or atmospheric scientists want any spare cycles... I have to draw a line somewhere don't I?
I bet if we do translate a message from ET it will say the following "Hey neighbor can you spare a planet!!!!!"
I dunno -- as inherently INTERESTING as the subject is, I suspect it'll be nothing more than a huge waste of processor cycles.
I mean, essentially we're starting out with random parameters (guided by what they offer as "realistic") and project forward with hypothetical rules on behavior to an unknown point in the future where we'll calculate the probability of a particular outcome?
Seriously, there's so much speculation and guesswork in even building the system to time-progress the ecological model that it seems unrealistic. Add on top of that the fact that no valid data at all will be used in the calculation, and of course the obvious limitation on the possibel number of factors we'll be calculating (as opposed to the huge number of factors that really exist) and you're pretty much pissing in the wind.
think about it this way -- assume everything in the simulation is perfect, including all the data it starts out with (two major assumptions) EXCEPT that a year from now we find out the reproductive cycle of the lungfish plays a greater than anticipated role in the production of kelp, which in turn is respoonsible for generating a large portion of the atmospheric oxygen. Every calculation done will be a complete waste of cycles and you'll have to start all over with the new ecological model.
or am i missing something?
Recursive: Adj. See Recursive.
In my (informed) opinion the overwelming consesus amongst climatologists and biogeographers is that climate change is real - and this is backed up by both modelled and experimental data - see Myneni et al, Nature 386 (1997) for some convincing evidence from our group.
Of course, you are perfectly welcome to download a GCM such as CCM3 and go through it line by line to see whether it is "real science" or not.
If you want to discuss what is "real science" or not email me or if you're in the Boston area, come round - my work address is on my web page.
Nick
-- "It's a sad day for American capitalism when a man can't fly a midget on a kite over Central Park" - Jim Moran
with so many of these projects (RC5, SETI, this, distributed.net) it seems sad that every time someone does a new distributed computing system they have to reinvent the wheel and make a totally new client from scratch.
would be nice if somone would just write a protocol and open-souce client for the darn thing. like, some way to send out abstract work blocks of various kinds over a network, processed with spare cycles, and have those work blocks returned.
it would have to be something very abstract, or at least extensible, so you could easily swap projects promiscuously without downloading and configuring a new client. and it would have to be processor-agnostic, maybe just put in a bunch of mathematic instructions (though a VM of any kind would just be stupid) and a standard way of parsing them. although you'd want to put in hints (vector instruction here, floating-point instruction here) so that things like MMX and altivec and 3d cards could be used to their full potential. And there would have to be _very_ clearly defined limits on the way they can access the hard drive, and ways to make sure it doesn't interfere with other applications. And it might need to try to make sure it only consumes network bandwidth if it isn't taking away bandwidth being used by something else. I dunno how you'd deal with the question of whether the work blocks are getting returned correctly; only thing i can think of is extreme redundancy. Send out all work blocks two or three times, if there's _any_ difference in the returned blocks redo it and maybe put the computer that returned the bad block on a list of computers not to trust. So it would be kinda complex to make a generic protocol instead of a specific implementation..
but wouldn't that be COOL?
especially if it wasn't just an internet thing, but a generic network thing; we have a _lot_ of computers at the school just sitting there all day waiting for someone to ctrl-alt-del and put a username in the login box. would be nice if they could be put to some meaningful use in their downtime.. like just say on one computer "rip this mp3 for me", or like an entire queue of mp3s and 3d renderings or whatever, and have all the computers on the network not in use do the work while i continue using the computer i'm on. 'course the network admin might not be too happy about his entire network being turned into an mp3 encoder, but hey, he doesn't need to know about it. It's his own damn fault for using NT, esp. without reading the damn manual..
distributed.net seems to be using the more use-a-specific-client-for-each-specific-task tactic, but maybe they could be raided for useful source code..
(p.s. mp3s are a hypothetical example, of course.. i wouldn't actually do that, that would be illegal! Riiight..)
-mcc-baka
INTELLECTUAL PROPERTY IS THEFT
Irritable, left-wing and possibly humorous bumper stickers and t-shirts
So, you get to use your "spare" cycles to help determine just how much the energy you're burning is going to damage the environment. Hmm. I'd like to see some estimates of which will do more good for the environment - participating in this program, or just turning your computer off.
We're above 4 trillion dollars and the computers designed to consistently add US$0.01 to the total were built back in 198x, so they must be having problems. A few more CPU cycles and our deficit counter can go up faster than ever!
Ever notice how Slashdot slows down sometimes? Ever think to yourself, "Man, if my spare CPU could go to speeding this bad puppy up"? Well, now you can speed up the experience with the only distributed client that actually sends CPU processing to Slashdot (warning: still in beta, no ETA).
Ever wonder how much consumable energy we have left? Well, with the World Resources Destructo-meter, you can help keep track of how many of our precious resources are left! As an added bonus, the more and more computers use this program, the more and more energy is used and thus, the program counts down faster and faster. It's fun for the whole family!
Ever given something you don't want to the Salvation Army? Ever wondered how anyone could use whatever it was you gave them? Now you can do the same thing with your CPU cycles! (NOTE: the Salvation Army is also selling used CPU cycles in their thrift stores for older computers).
NASA may not seem like it needs more CPU cycles, but, as current events recently showed, the CPU-intensive conversions from standard to metric and back again sometimes hit a brick wall. Donate CPU to NASA and make sure that we don't lose anymore multi-million dollar probes!
Brought to you by The Computer(tm), now with new games that you play when you're not using your computer.
While it make be hot and sexy to participate in a feel-good global climate project, I'd like to see more of the scientific methodology first. Anyone can run monte-carlo simulations (in fact too many do so for market forecasting) but the underlying science still has to be validated. While it may save them buying a Cray or 10, will it achieve any useful results? I recall a project which simulated the effect of climate change across a forested German country-side and what eventually hit the papers bigtime was the most extreme scenario where all the trees were killed off. Essemble forecasting can usually pick the extreme events but the normal one are trickier to analyse. Extrapolate this across the globe and you will always find a scenario where you are personally affected (wheat belts moving into Canada, sea levels rising, higher winter fuel costs, etc). There is a very good reason why science has to go through peer review first and to be especially sceptical of simulated results which makes a lot of simplifying assumptions (which would be necessary to fit into a PC memory).
So what will the climate be like next decade? All anyone can say really say is that it might change. Any attempts at scare-mongering or trying to protect vested interests will only be a distraction to putting in the fundamental research in trying to gain a better understanding of the world's climate dynamics.
LL
- SETI@home : a neat, geek-friendly, worthy cause, but hardly practical, and they have plenty of CPU - they don't need mine. This project is currently running, and while they had growing pains a while ago, they have been solved. One problem remains (as far as I know); they client still slows to 1/3 speed if you have the visuals turned on.
- distributed.net encryption cracks : for a long time, the most practical distributed computing projects around, and certainly the most advanced, but I think that they (and others) have clearly proven the point re: encryption (i.e. that we need access to stronger encryption), and while we don't have perfect regulations, one more crack ain't gonna do it. These are also running, and have been for years now. They had some problems with stats a while ago, but they haven't had a problem of not sending out new work in a long time (if ever).
- d.net's OGR project : good scientific research, certainly, but really, just giving some grad student thesis fodder. You're not gonna solve world hunger or anything. Also, it isn't running yet.
- Casino-21 Project : wow, some practical application. Also very different from other projects, because it is devoted not to "solving a riddle", but to predicting a complex system. Of course, it's not running yet, either.
I'm sure there are others, but that should cover the big ones, and all of the major categories. For now, it's d.net rc5-64 for me, but as soon as Casino-21 starts, I'm going to switch. I hate to sound like a whiny earth-tone, but I have to say I think the environment is important, and certainly is much more proximate to me than aliens (I think) (I hope, too). This project may not solve global warming, but as GI Joe says, "Knowing is half the battle."-Yoshi
Take a gander at John Daily's page, Waiting for Global Warming, from where you can follow links to NOAA and NASA evidence suggesting that the global warming phenomenon is really systematic error in the climate record.
But by all means, if you don't want to look for ET reruns of "I Love Lucy" analogues, then pitch in on a decent atmospheric model. If we're actually headed for an Ice Age, it'd be nice to know in advance.
I can see the fnords!
This shows only that projects like GIMPS (which IMHO beats distributed.net), distributed.net and SETI@Home has done what part their original intention is: Lead the way for other distributed efforts.
:-)
Without having looked at the article, though, it looks like this will be much less accessible for the masses. GIMPSers (check http://www.mersenne.org if you're interested) have months or even years to complete a single assignments -- this one sounds like it will need the data in time. In other words, only 24/7 (or almost-24/7)-online computers will be able to participate effectively. I wonder how it will cope with having a variable amount of information available as well. OK, off to read the article
/* Steinar */
(This comment is of course GPLed.)
This isn't a case of donating spare cycles to a project. This is a case of donating spare COMPUTERS.
Take a quick look at the registration form. The minimal machine they're interested has a 400Mhz CPU, 32MB RAM, and 256MB disk space they can devote to the task. They'd prefer that you have a 600Mhz machine with over 2GB to devote. They say themselves that the system requirements are about equivalent to a state-of-the-art game... a game you'll be running for about a year.
Unlike distributed.net, I'll bet this is going to have serious repercussions on your usage of your own computer. (After all, it'll take quite a bit to swap out when you come back to your machine.) And what do you get for this? Well, in 50 years we'll know if your random simulation was better than any random simulations. At which point...
If you'd like to participate in this, you'd might as well fedex your machine to them so that they can put them on a high-speed network and beowulf them. With the kind of data they need, the network latency is going to be bad... (distributing things by CD?!)