My computer is working. Is yours?

[amethyst73]

Yesterday evening I caught sight of this headline over at Google News, about PS3s being used in distributed computing problems.  It reminded me about a project that the huz and I had been slightly involved in some time ago, largely due to his PhD work in simulating protein folding.  Vijay Pande's lab over at Stanford University has been running the Folding@Home project for some years now.  The basic idea is that simulating a protein folding over any reasonable length of time with any reasonable amount of accuracy takes a huge amount of processing power.  The folks at Folding@Home are running a system where small bits of the processing are farmed out to many many many individual computers all over the world.  The individual computers then send back the results of their itty bitty bits of processing to the main computers back at the Pande lab, and the lab puts all the results together.  It's a pretty neat project.

Anyway, it occurred to me that my computer at work is on all day, largely sitting and running a screensaver (if it's not doing something terribly exciting such as running an Excel spreadsheet - that's most of the work-related business that this computer does!), and that I could very easily put it to work doing something that's actually useful to someone.  So I've downloaded the graphical interface module and am currently watching a model of supervillin bounce around.

If you're more into space stuff than proteins, there's also the SETI@Home project, which I believe was actually first on the block to come up with the massively-distributed-computing idea.  It distributes analysis of radiotelescope data in the search for extraterrestrial intelligence.  I probably wouldn't get into any trouble with running that at work as well, but I'd rather be running something which is a bit more closely related to what I actually work on.

If you've got some computing power to spare, do consider contributing to one of these projects.  If you start it running before you leave for the weekend, who knows what your computer might fold or find by the time you come back on Monday?  :)

Comments

[haamel.livejournal.com]

This type of application goes by many names, including "cycle sharing". Many industrial settings (including mine) exploit this in-house to farm batch jobs to underutilized servers and workstations. The upside to tapping all that unused computing power is obvious.

There are a couple drawbacks, however. Network bandwidth can be an issue, if the problems being farmed out cannot be exchanged concisely, or if communications must take place too often. A more insidious issue is power consumption, since modern CPUs play an increasing array of tricks to reign in their *average* power consumption. Applications like Folding@Home directly translate to energy consumption (and energy generation byproducts like "carbon") that have a dollar value if one digs far enough.

I don't say that cycle sharing is a bad idea for society, but I do advocate a balanced grasp of the tradeoffs it implies. The amount of additional energy bill I'd accept for the good of SCIENCE(tm) is surely more than 1%, but surely less than 50%. ^^;

[amethyst73.livejournal.com]

A more insidious issue is power consumption, since modern CPUs play an increasing array of tricks to reign in their *average* power consumption. Applications like Folding@Home directly translate to energy consumption (and energy generation byproducts like "carbon") that have a dollar value if one digs far enough.

Interesting! Thanks for pointing out this issue.

Do you happen to know where one could go to find out how much/what fraction of computer power one is using by running cycle sharing programs in general? (I'll cheerfully admit I haven't looked carefully at Folding@Home's page.)

...At present, my Activity Monitor says that Folding is taking up roughly the same percent of CPU time as my network browser (between 5 and 10 at most updates - and now that I'm typing in my browser, it's taking up between 10 and 20%).

[haamel.livejournal.com]

Well, the calculation Folding@Home lists on their website is here: http://folding.stanford.edu/faq.html#misc.power. It would be interesting to see if Intel (who is listed as one of the apparent sponsors of the project) has anything official to say on the topic. I can try to track someone down to ask at work.

[karakara98.livejournal.com]

I was going to bring up this issue too.

So if you computer is going to be on all day anyway, running a cycle sharing program may not add much to the power consumption. However, you would likely save more power and more carbon by puting the computer to sleep during the day and shutting it down at night.

I'll look around at work to see if I can find any data on computer power consumption, which shouldn't be too hard since I'm now in the energy efficiency industry. My gut instinct is that the social optimum is running the cycle sharing when your computer is on but not in use during the day and shutting the computer down at night and on the weekend.

[amethyst73.livejournal.com]

Thanks for the link!

I chatted with the huz about this issue too, and it seems likely that there's another issue too: how much processing you're doing per unit energy. I have one of the 'old' lampshade-style iMacs at work with an 800MHz PowerPC G4 chip. That's slow enough that I may not be getting a terribly good amount of calculation per unit of energy that my employer is paying for to keep my computer running - ergo, not so much calculation per unit carbon.

In my particular case, overall benefit might be increased if I just told my computer to put the CPU to sleep after 15 minutes of non-use (it's currently set to not sleep, which I stupidly realized only the other day!)

[amethyst73.livejournal.com]

Just posted a sort-of detailed response to haamel's newest comment. Short story: Old chip, slow processor (800 MHz G4). Society overall might well be better off if I actually told the thing to go to sleep after 15 minutes of non-use!