Back to previous post: “Regression toward a phony mean”

Go to Making Light's front page.

Forward to next post: Clean Freak Confessions

Subscribe (via RSS) to this post's comment thread. (What does this mean? Here's a quick introduction.)

April 20, 2009

Wide Area Computing
Posted by Jim Macdonald at 09:40 AM *

Y’all remember SETI@Home, right? That’s a project that’s been going on for years: people using their home computers as part of a large distributed project to search for signs of intelligence in the radio signals received at Arecibo. The government won’t fund it (and we can all understand why), but there’s no reason that others might not be interested.

The way SETI@Home worked was you installed a screen saver on your computer. It would download raw data, and process it while you weren’t using your machine, then upload the job and get more data. Meanwhile, it put some pretty graphics on screen.

Well, this grew, and eventually SETI@Home moved to a new location with new software. This is pretty, and now there are more projects (just in case you think that SETI is a waste of time).

What we have now is Berkley open infrastructure for network computing (BOINC). It’s still a screen saver (not that screens need much saving these days), and it still uses idle time on your CPU to create a huge virtual CPU for really big computing projects. More than that, you can have multiple projects, and it’ll cycle among ‘em. (You can also join teams and do other stuff. I’m on one of the SFF Net teams for SETI, and I joined the Navy team for another project.)

Let’s see: what projects do I have running here?

  • SETI@Home. The classic. Search for Extra-Terrestrial Intelligence.
  • Einstein@Home Searchs for spinning neutron stars (pulsars) using data from the LIGO gravitational wave detector.
  • Rosetta@Home. As you know, Bob, proteins have shapes, and the shapes determine what they do. One chemical formula can have many shapes. Rosetta is trying to determine the 3-dimensional shapes of proteins in research that may ultimately lead to finding cures for some major human diseases.
  • Climateprediction.net. This one is to produce predictions of the Earth’s climate up to 2080 and to test the accuracy of climate models.

Some other projects.

BOINC System requirements. (Versions available for Windows, Mac, and Linux.)

Comments on Wide Area Computing:
#1 ::: Alice ::: (view all by) ::: April 20, 2009, 10:52 AM:

I'm surprised you're not also connected to Folding@home, what with the medical research aspect... http://folding.stanford.edu/

#2 ::: James D. Macdonald ::: (view all by) ::: April 20, 2009, 11:24 AM:

I was doing both SETI@Home and Folding@Home under the old software. When the software changed I went with Rosetta@Home instead, since it was doing basically the same thing.

#3 ::: Cindy L. ::: (view all by) ::: April 20, 2009, 11:26 AM:

Thanks for sharing! I just signed my work computer up for Rosetta. I figure that it should be working on proteins while I'm working on DNA out in the lab.

#4 ::: TomB ::: (view all by) ::: April 20, 2009, 11:27 AM:

Folding@home is a good project but it doesn't show up on the BOINC list because they wrote their own client. Stanford competing with Berkeley, who'd a thunk it.

#5 ::: Liza ::: (view all by) ::: April 20, 2009, 12:14 PM:

I'm intrigued by these things (and have been for some time) but I'm nervous about allowing remote access. In y'all's opinion, is it really safe?

#6 ::: Bill ::: (view all by) ::: April 20, 2009, 12:20 PM:

I used to run GIMPS, the Great Internet Mersenne Prime Search, and for a while I was running Folding@Home. I'd recommend against running any of these CPU-burners on laptops, unless you're really careful about your batteries and machine temperature - laptop battery systems are designed for low average CPU use and don't like the kind of heavy drain and frequent deep cycling they'd get if you run it on an hour's train commute twice a day, and laptops tend to overheat easily.


On the other hand, if you have a game console, or have one of the higher-end graphics boards in your desktop PC, there are versions of many of the popular clients that will run an order of magnitude faster on the graphics processor than on a typical PC CPU, and can kick out results a lot faster than my old desktop.

#7 ::: Bill ::: (view all by) ::: April 20, 2009, 12:24 PM:

Liza#5, remote access isn't a risky thing here - the standard systems aren't acting as a server waiting for some central system to connect to them, they're a client that fetches new work from the central server when the client needs it, using HTTP or some similar data-requesting program. Obviously you should only run software from trusted sources, yadda yadda, but assuming you do that, this is a much safer environment than your web browser.

#8 ::: Madeline Ashby ::: (view all by) ::: April 20, 2009, 12:24 PM:

I've always been fascinated by these distributed projects, but now I wonder what the benefit of spreading them across home networks is when you can download Linux onto a cluster of PS3's and start running simulations on a private supercomputer. The distributed project might still be cheaper (you need about eight PS3's to make the magic happen, but sixteen is better), but you'd have more control over your own cluster. Am I just wildly underestimating the computational power necessary for these projects? (Probably.)

#9 ::: KeithS ::: (view all by) ::: April 20, 2009, 12:31 PM:

Liza @ 5:

It looks like the BOINC system communicates over the internet using HTTP, so you shouldn't need to poke any holes in your firewall to let it run. If you get a message from your firewall, that's likely to be one asking if the program should be allowed to access the network at all, which you want it to do.

It appears that it's only initiating connections with the server, not listening for incoming connections, so it should be pretty safe.

#10 ::: Doug K ::: (view all by) ::: April 20, 2009, 12:33 PM:

I was running Climateprediction a couple of years ago, but they switched to Boinc, which was then maddeningly unstable: gave up.

Started with The Clean Energy Project a few months back, which also uses Boinc; but fortunately it no longer crashes my PC and/or throws away the computed results.

#11 ::: Clifton Royston ::: (view all by) ::: April 20, 2009, 12:39 PM:

So what you're saying here is "Scientific progress goes BOINC"?

#12 ::: Earl Cooley III ::: (view all by) ::: April 20, 2009, 01:16 PM:

I think it will be only a matter of time before someone writes an Astroturf@Home system that constructs political troll forum posts like unto combining amino acids in the semblance of life itself....

#13 ::: Linkmeister ::: (view all by) ::: April 20, 2009, 01:53 PM:

Liza @ #5, I've been running both Rosetta and SETI for years, with Zone Alarm's firewall and with others. I've never had an issue with either of them.

I like looking at the Rosetta screen saver; trouble is, my monitor insists on going to sleep after a little while, so even screen savers don't stay visible for long.

#14 ::: Chris Eagle ::: (view all by) ::: April 20, 2009, 01:59 PM:

@8: You're underestimating. Folding@home is running on about four hundred thousand machines, including over fifty thousand PS3s.

#15 ::: Madeline Ashby ::: (view all by) ::: April 20, 2009, 02:09 PM:

@14: Thank you! I figured I was wrong about the scale.

#16 ::: rm ::: (view all by) ::: April 20, 2009, 02:23 PM:

I confess that the acronym BOINC, plus the Particle about precision-hacking the TIME 100 online poll which emerges from a subculture alien enough to me that I have trouble crediting it as real, made me suspicious. I checked to see if it's April 1st again. But I suppose that these are phenomena of the real world, which only seem like particularly parodistic sf scenarios.

#17 ::: David Dyer-Bennet ::: (view all by) ::: April 20, 2009, 02:36 PM:

Been running SETI@Home since far pre-BOINC, both on my desktop and on server systems at home. Anybody with Minneapolis fannish connections should consider joining the SETI@Home Minn-StF team :-). For a while our most prolific contributors were in California, but I think Minnesota has gained back the lead within the team.

Running these things mostly makes sense if your computer is on full-time. And do remember that being fully on rather than in power-save mode will make a difference to your electric bill.

#18 ::: Brooks Moses ::: (view all by) ::: April 20, 2009, 03:27 PM:

David @17: Indeed; that's one of the big changes since SETI@Home started -- computers have become much better about not having "spare CPU cycles" in the original sense. It used to be that (to a reasonable approximation) a computer that was only sitting and waiting for you to press a key would be using just as much power as if it were actively computing as much as it could, so there really was no additional cost to running these things. These days, a computer that's not doing computations has all sorts of ways to reduce its power consumption by quite a lot, especially if you're talking about computations on the graphics card as well as the CPU, and so the unused cycles aren't being wasted in the same way. Moreover, these days the cost of a computer compared to the cost of the power required to run it is going down, so that pushes the balance even farther towards the incremental cost of running these programs instead of the already-sunk fixed cost of the idle hardware.

It's still valuable work, but it's not nearly the "something for nothing" that it once was. At a rough guess, I'd say the costs of running SETI@Home on four hundred thousand machines instead of letting those machines sit idle are somewhere in the millions of dollars per month.

(The fact that something like this can get people to contribute that much cash, via their electric bill, to such a project is itself quite a remarkable thing, IMO.)

#19 ::: mcz ::: (view all by) ::: April 20, 2009, 03:30 PM:

I'll put in the good word for the World Community Grid which runs several humanitarian projects (cancer, dengue, rice, AIDS, clean energy, proteome folding).

The research is conducted by public or non-profit orgs, and results must be released to the public domain.

#20 ::: Liza ::: (view all by) ::: April 20, 2009, 03:34 PM:

Thanks, everyone, for the reassurances! And DDB, yes, if I run SETI@Home I'll join the MN-StF team. :-)

#21 ::: Michael Roberts ::: (view all by) ::: April 20, 2009, 03:36 PM:

I would like to mention at this point that the Arecibo telescope is very, very cool when visited in person. Plus they have the best science museum on the island. (Well, OK, the only science museum on the island that I know of -- but it's a good one.)

#22 ::: --E ::: (view all by) ::: April 20, 2009, 03:37 PM:

Unexpected beneficial side effect of this program: it can find a stolen laptop.

True story: a friend of mine had her laptop stolen. The thief used it as it was, not wiping the hard drive. A couple of days later, it called in to SETI@Home, and the police were able to locate it by the IP address.

#23 ::: Bill ::: (view all by) ::: April 20, 2009, 05:40 PM:

For several years, when SETI@Home was the main distributed processing application running and when its web page made it easy to see how fast it was going, SETI was always significantly faster than any of the machines on the Top500.org list. You can't really directly compare the FLOPS ratings of a supercomputer running LINPACK with a loosely-coupled network running designed-for-easy-parallelism applications, but it was still gratifying that regular people looking for space aliens in their spare time had a supercomputer several times faster than the US nuclear weapons labs had built to improve weapons of mass destruction.

Some years the fastest supercomputers have been non-evil (The NEC Earth Simulator and other weather prediction machines need really heavy processing, and more recently some of the IBM BlueGene machines may be doing protein folding, though since they're at the weapons labs that may not be the case.) There's been a really radical increase in performance the last few years, largely driven by technology that was developed not for NASA or the military, but for the Playstation 3.


The current total speed of BOINC is about 1.1 PetaFLOPS, similar to the fastest machine on the Top500 List. Folding at Home is at about 5 PetaFLOPS these days - it took a big jump two years ago when the Playstation 3 client came out. On the other hand, the nuke labs do have a bunch of machines in the Top500 (and Top50) class, so they've got more total horsepower than we do. BOINC's largest application is still SETI (looked like almost half of their horsepower), but they and Folding both have a wide variety of research these days.

#24 ::: Bruce Cohen (SpeakerToManagers) ::: (view all by) ::: April 20, 2009, 07:49 PM:

I used to run SETI@Home, but my personal computer is now a laptop, and I don't get the advertised battery life anyway, so I've stopped doing that. Someday, when either I have a desktop again or batteries get better, I'll go back to running that sort of software; it's a neat way to be a part of modern science.

By the bye, the results from protein folding software will probably be extremely useful for early nanotech development; it's looking now like the initial phase of assembling nanomechanical and nanoelectronic components will be done with assembly tools made of proteins, picking up atoms or small molecules for assembly with their active sites. See the archives of Eric Drexler's blog, MetaModern for several articles on using biomimetic techniques like protein catalysis and macromolecule self-assembly for nanomechanical synthesis.

#25 ::: Raphael ::: (view all by) ::: April 21, 2009, 05:14 AM:

Can't this stuff "wear out" your HD, and perhaps your CPU or other hardware, faster?

#26 ::: Steve Burnap ::: (view all by) ::: April 22, 2009, 10:37 AM:

I ran the PS3 Folding@Home for a while, but the hardware really sucks down the power. It was literally increasing my power bull by $10/month. (Admittedly in Summer, so I was at the highest PG+E tier with the air conditioning.)

I work for Sony, and am very happy that we are making it so easy for people to contribute to this thing. (Even if I don't run it myself very often.)

#27 ::: Wyman Cooke ::: (view all by) ::: April 25, 2009, 08:29 PM:

I had SETI@Home on a couple of computers. I haven't kept it going since I got the laptops.

#28 ::: MichaelJ ::: (view all by) ::: April 26, 2009, 05:21 AM:

Open question: any thoughts on whether to prefer Folding@home or BOINC? I'm a student worker for a university tech service, and I'd like to propose rolling out one of these to our computer labs over the summer. But I haven't been able to decide which to suggest.

#29 ::: Xopher sees spam ::: (view all by) ::: January 22, 2010, 12:03 PM:

Old thread. Generic comment. URL for irrelevant article.

#30 ::: Serge eyes SPAM ::: (view all by) ::: January 22, 2010, 12:03 PM:

dry ocular orbs

#31 ::: James D. Macdonald ::: (view all by) ::: January 22, 2010, 12:05 PM:

Xopher, I'm already on it. It just takes me a minute....

#32 ::: James D. Macdonald ::: (view all by) ::: January 22, 2010, 12:41 PM:

To answer MichaelJ's question from last spring (sorry about the delay!) -- I've tried both Folding@home and BOINC, and I've gone with BOINC on the latest computer, purely because of the wide variety of other projects that can be slotted in.

Choose:
Smaller type (our default)
Larger type
Even larger type, with serifs

Dire legal notice
Making Light copyright 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017 by Patrick & Teresa Nielsen Hayden. All rights reserved.