SETI@home questions and options
April 23, 2009 9:34 AM Subscribe
For a while I have been using the SETI@home screensaver. I do not know much about computer science but my understanding is that the purpose of the program is to create a virtual super computer by breaking up packets of information and then sending them out to all of these computers all over the world which then run them through the program and send back the information. So you end up having the same computing power as a super computer through a large number of home computers. I think this is a fascinating concept but I have 3 questions:
1. How much additional energy does this utilize? I mean is my computer now using the same amount of energy as when it is normally running?
2. Do limits to this system exist? IE what can this kind of system not do that a super computer could?
3. Anyone know of any other programs like this which allow your computer while you are not using it to be searching for knowledge?
1. BOINC Heat and energy considerations.
2. Grids versus conventional supercomputers
3. List of distributed computing projects.
posted by zamboni at 9:41 AM on April 23, 2009 [3 favorites]
2. Grids versus conventional supercomputers
3. List of distributed computing projects.
posted by zamboni at 9:41 AM on April 23, 2009 [3 favorites]
1. A lot. SETI will max out your computer's CPU which draws much more power and generates more heat than when your computer is idle. Completely wild guess: 10x the power consumption of your computer at idle, 100x the consumption of your computer when it's asleep.
2. distributed computation projects like this are best where the problem is 'embarrassingly parallel'. Each computer can work on its part with no communication and minimal coordination with other computers in the network. Two classes of problem which do not fit this model well: sorting, fourier transforms.
3. See zamboni's link above.
posted by zippy at 9:47 AM on April 23, 2009
2. distributed computation projects like this are best where the problem is 'embarrassingly parallel'. Each computer can work on its part with no communication and minimal coordination with other computers in the network. Two classes of problem which do not fit this model well: sorting, fourier transforms.
3. See zamboni's link above.
posted by zippy at 9:47 AM on April 23, 2009
The wikipedia article on Parallel computing is very relevant, especially if you really want to dig into the technical details.
posted by chrisamiller at 9:51 AM on April 23, 2009
posted by chrisamiller at 9:51 AM on April 23, 2009
BOINC, as mentioned above, is a platform for many different projects to use the same basic distribution method. This allows projects to focus on their specific needs, while leaving the mundane tasks to the BOINC team. It's also handy in that you can attach to multiple projects - if one runs out of work temporarily, or goes down, your computer will work on other projects in the meantime.
SETI@home used to be its own project, but eventually became a part of BOINC. I usually keep 2-3 projects going on my PC at any given time, while my PC is idle (Usually while I am not there).
posted by GJSchaller at 9:59 AM on April 23, 2009
SETI@home used to be its own project, but eventually became a part of BOINC. I usually keep 2-3 projects going on my PC at any given time, while my PC is idle (Usually while I am not there).
posted by GJSchaller at 9:59 AM on April 23, 2009
One benefit traditional supercomputers have over grid computing is the lower communication latency (fancy way of saying communication is quicker between processors.)
Some problem sets, like SETI, or folding can easily be broken into chunks that can be processed without any interaction amongst the processors. These types of problems are commonly referred to as "embarrassingly parallel" if you wish to google/wiki more.
other types of problems (for example: physics simulations involving particle interaction) require communication, so a traditional supercomputer, or a "cluster" (the poor man's super computer built from many regular computers in a single high speed network) is a more attractive platform. This isn't a limitation of raw power, but of design trade-offs.
posted by oblio_one at 10:03 AM on April 23, 2009
Some problem sets, like SETI, or folding can easily be broken into chunks that can be processed without any interaction amongst the processors. These types of problems are commonly referred to as "embarrassingly parallel" if you wish to google/wiki more.
other types of problems (for example: physics simulations involving particle interaction) require communication, so a traditional supercomputer, or a "cluster" (the poor man's super computer built from many regular computers in a single high speed network) is a more attractive platform. This isn't a limitation of raw power, but of design trade-offs.
posted by oblio_one at 10:03 AM on April 23, 2009
1. How much additional energy does this utilize? I mean is my computer now using the same amount of energy as when it is normally running?
It actually probably uses more energy than when the computer is normally running with you using it, assuming SETI's using the processor(s) at or near 100%, where your normal computer usage is processor light stuff (web browsing, listening to music, word processing, etc). Just poking around a ballpark figure for a high performance system might be around twice the overall power usage when the processors are maxed vs. the system idling, but that's just a really rough guess. If you're concerned or just interested pick up a Kill-a-Watt, which will monitor power usage for any appliance you hook it up to. Pretty handy to have, IMO.
posted by 6550 at 10:25 AM on April 23, 2009
It actually probably uses more energy than when the computer is normally running with you using it, assuming SETI's using the processor(s) at or near 100%, where your normal computer usage is processor light stuff (web browsing, listening to music, word processing, etc). Just poking around a ballpark figure for a high performance system might be around twice the overall power usage when the processors are maxed vs. the system idling, but that's just a really rough guess. If you're concerned or just interested pick up a Kill-a-Watt, which will monitor power usage for any appliance you hook it up to. Pretty handy to have, IMO.
posted by 6550 at 10:25 AM on April 23, 2009
Probably out of date now but a while back I wrote an article about seti@home and grid computing, might be of some use.
posted by fearfulsymmetry at 12:02 PM on April 23, 2009
posted by fearfulsymmetry at 12:02 PM on April 23, 2009
There is a great paper about the economics of distributed computing which focuses on Seti@Home as a case study. It was written by Jim Gray back in 2003.
Distributed Computing Economics
(and here's an (html version)
But this stuff is out of date now, yes. I think Seti and similar projects are doomed because people are conscious about the energy their computing systems are using, and they recognize that these spare cycles aren't "free" - they have a monetary cost because they keep the system running when it could be idle or sleeping.
posted by PercussivePaul at 1:12 PM on April 23, 2009
Distributed Computing Economics
(and here's an (html version)
But this stuff is out of date now, yes. I think Seti and similar projects are doomed because people are conscious about the energy their computing systems are using, and they recognize that these spare cycles aren't "free" - they have a monetary cost because they keep the system running when it could be idle or sleeping.
posted by PercussivePaul at 1:12 PM on April 23, 2009
That's kind of a bummer, the screensaver program running the processor at 100%, using lots of energy, heating up the CPU, the power supply, and the room the computer is in.
Is there any way to manually throttle the app or the CPU to slow it down and cut back on energy usage?
posted by exphysicist345 at 10:55 PM on April 23, 2009
Is there any way to manually throttle the app or the CPU to slow it down and cut back on energy usage?
posted by exphysicist345 at 10:55 PM on April 23, 2009
The real limit is the fact that unless the problem can be broken up into little bits then its a bad candidate for distributed computing. A real supercomputer would have very quick communications between its nodes and could, say, analyze real world weather data, but realtime processing via distributed computing would be difficult to say the least.
Another thing to think about is that SETI is something of a pipe dream. For instance, we coudnt detect a civilization like ours. Nor are we broadcasting directly at systems assumed to have lfe so they can detect us. They would need to be much more advanced than us to even detect it. Toss in some of the factors of the Drake equation and all of this seems like a waste of energy.
You can also donate your computer time to research problems for science. Its all a trade off, if 1 million computers are running when they normally wouldnt be then thats 10x more power used. If that much power just means more coal being burned and more pollutants in the air, then you need to ask yourself if any of this is worth it. I think its important to consider the holistic effect here and how we make energy.
posted by damn dirty ape at 11:09 PM on April 23, 2009
Another thing to think about is that SETI is something of a pipe dream. For instance, we coudnt detect a civilization like ours. Nor are we broadcasting directly at systems assumed to have lfe so they can detect us. They would need to be much more advanced than us to even detect it. Toss in some of the factors of the Drake equation and all of this seems like a waste of energy.
You can also donate your computer time to research problems for science. Its all a trade off, if 1 million computers are running when they normally wouldnt be then thats 10x more power used. If that much power just means more coal being burned and more pollutants in the air, then you need to ask yourself if any of this is worth it. I think its important to consider the holistic effect here and how we make energy.
posted by damn dirty ape at 11:09 PM on April 23, 2009
You can also donate your computer time to research problems for science. Its all a trade off, if 1 million computers are running when they normally wouldnt be then thats 10x more power used. If that much power just means more coal being burned and more pollutants in the air, then you need to ask yourself if any of this is worth it.
I think there's an environmental argument the other way too. These projects need to get done, so the scientists can either buy a cluster or use distributed computing. If they buy a cluster, you have to consider the impact of all the extra manufacturing, toxic metals in the motherboards, etc. By using existing, underutilized resources, they save all the CO2 that creating and shipping the cluster parts would release.
That said, it would be great if we could work out some sort of micropayment system for this stuff. Give the people running the programs a few bucks for their services, or better yet, let them deduct that money from their taxes. In essence, those spare cycles and electricity are a charitable donation, and free up more NSF/NIH money for other projects.
posted by chrisamiller at 1:18 PM on April 24, 2009
I think there's an environmental argument the other way too. These projects need to get done, so the scientists can either buy a cluster or use distributed computing. If they buy a cluster, you have to consider the impact of all the extra manufacturing, toxic metals in the motherboards, etc. By using existing, underutilized resources, they save all the CO2 that creating and shipping the cluster parts would release.
That said, it would be great if we could work out some sort of micropayment system for this stuff. Give the people running the programs a few bucks for their services, or better yet, let them deduct that money from their taxes. In essence, those spare cycles and electricity are a charitable donation, and free up more NSF/NIH money for other projects.
posted by chrisamiller at 1:18 PM on April 24, 2009
Thats true but I bet if no one volunteered to run SETI then it would fold or at least run on much less equipment. Or that prime number program. There's a lot of optional wasted energy out there.
There's also the argument of locality. Id rather support something in France, which gets most of its energy from nuclear, than in the US where we burn coal.
posted by damn dirty ape at 2:15 PM on April 24, 2009
There's also the argument of locality. Id rather support something in France, which gets most of its energy from nuclear, than in the US where we burn coal.
posted by damn dirty ape at 2:15 PM on April 24, 2009
This thread is closed to new comments.
posted by Leon at 9:41 AM on April 23, 2009