SETI@home Becomes Part of BOINC 184
Sudoku writes "On December 15th the Seti@home project will stop issuing new work to members and integrate with BOINC, the Berkeley Open Infrastructure for Network Computing. Once members have moved over to the BOINC client they can divide their computing time between such projects as climate prediction, search for gravitational signals emitted by pulsars and yes, you can still look for the aliens."
BOINC blows (Score:4, Interesting)
Lose members (Score:4, Interesting)
More Practical Matters (Score:5, Interesting)
Still...won't be quite the same as when some guys in my last job rigged another fellow's screen saver to flash that his computer had found an alien signal.
sigh
BOINC could be a lot more efficient (Score:5, Interesting)
The following numbers are synthetic: I chose them to make the math easy. Let's say there are two distributed computing projects to choose from: OGR and RC5. There are also two different computers you can use to work on the projects, a G5 or a P4.
The G5 can complete 3000 units of OGR in one hour and 1500 units of RC5
The P4 can complete 1500 units of OGR in one hour and 1000 units of RC5.
I have a P4 and like to work on OGR, while my friend Eliza has a G5 and prefers to work on RC5. We each fire up our distributed clients and let them run for two hours, then check our stats:
OGR on P4: 2 hours * 1500 units/hour = 3000 OGR units
RC5 on G5: 2 hours * 1500 units/hour = 3000 RC5 units
Now let's see what comparative advantage has to offer. The P4's ratio of efficiencies is 1500 OGR units/hour to 1000 RC5 units/hour, or 3 OGR/2 RC5. The G5's ratio is 2 OGR/1 RC5. In other words, even though the G5 is better at both OGR and RC5, it is relatively better at OGR.
I already know I can crunch 3000 OGR units in two hours. Instead of actually doing this, I ask Eliza to work on OGR for me while I do RC5 for her. Now what happens?
OGR on G5: 2 hours * 3000 units/hour = 6000 OGR units
RC5 on P4: 2 hours * 1000 units/hour = 2000 RC5 units
This is great for me, 6000 OGR units were completed. But Eliza's not happy because the RC5 work is falling behind. What happens if she works on each project for an hour while I work on OGR for
This shouldn't be too difficult to implement. With BOINC, instead of choosing which project their computer will actually work on, a user submits their project preferences. Then the client runs a series of benchmarks that determine the computer's ratios of efficiencies. These data are sent to the distributed server which determines the optimal allocation of work between all clients, while guaranteeing each client that as much or more work will be done on the project of their choice as would occur if that client worked solely on its preferred project.
Someone should invent (Score:2, Interesting)
Re:BOINC blows (Score:3, Interesting)
and i've been using this boinc thingy for like months now to run my seti
otherwise the cpu time sharing between different tasks and etc. is a good idea , thumbs up, but for the complexity, thumbs down.
Re:Foldit (Score:4, Interesting)
-Rick
Re:Odds of finding aliens (Score:3, Interesting)
But still, they come.
Re:BOINC blows (Score:5, Interesting)
The news is that old Seti is finally dying, and not in the silly "netcraft confirms" way, but finally going away.
The comments about the move over the few threads that have talked about it are freaking hilarious. I've never seen so many (reasonably) tech savvy people turn into 85 year old codgers. "My Opteron processes 14 Seti@home classic units to the hogshead, and that's the way I likes it!"
Seti Classic hasn't been doing anything productive for *years*. The work units you were running were validations and revalidations of already validated workunits. You may as well have created 500 blank word documents and set up a windows task to copy them from one partition to another for all the good it was doing. The "real" work was moved over to BOINC long ago. Classic is dead, remember how cool it was, and move on.
What I can't figure out is how people are having problems figuring out BIONC. Download BIONC. Install. Sign up with whatever @Home project you're interested in using. Go back to BOINC, attach to project using account key that was e-mailed to you (or e-mail address.). Walk away and wait for client to it's thing. Sometimes, especially during
Seti seems to have taken care of their last few bottlenecks, and opening up the old servers to start doing something useful should take care of the rest of any other capacity issues they've had. BOINC has been a huge improvement over classic. You can't fake results to run up your "score", the client is much more responsable when it runs out of work (trying to reconnect at growing random intervals during an outage, instead of constantly hammering away like a screaming toddler), and workunit queueing is handled within BOINC instead of through a third party system. I kinda miss the command line client, but that's about it.
Re:BOINC blows (Score:4, Interesting)
When I first started this, about a month after the project went public, I thought maybe it might be worthwhile. But now that I see the data is from a rather narrow band around the ecliptic and not from the whole sky, I'm not so sure we'll find anything in the more sterile environs of the milky way. To much sterilizing radiation down in the inner core for anything to have time to grow into something we might want to meet in between supernovas. Something we might not want to meet, maybe...
--
Cheers, Gene
Patent policy? (Score:3, Interesting)
Folding@Home at least say that they are a nonprofit and will not profit from selling or licensing their research.