Slashdot Mirror
BOINC Project to Search for Gravitational Waves
Buzz Skyline writes "Einstein@Home is a new, BOINC-based distributed computing project that will analyze data from the Laser Interferometer Gravitational wave Observatory (LIGO). The goal is to perform a whole-sky, gravitational wave survey of pulsars. Beta-test versions of the Einstein@Home screen saver should be available by the end of the summer, and final release is planned for early 2005."
Boinc should open up more distributed computing projects as well, since the server/client infrastructure is mostly prewritten. Since my other Boinc projects have been sputtering and not giving me work lately, maybe I'll give this one a try. More info on Boinc Here
Even those who arrange and design shrubberies are under considerable economic stress at this period in history.
For one thing, on most of the workstations BOINC would appear to work very quickly on the data only to crash out well before the computation was created. Indeed, sometimes it would actually crash before any data was processed by the application. At other points it would work for hours and hours without actually achieving anything; closing down the workstations at the end of the day without getting one computed dataset off was quite frustrating. On the workstations that were actually computing datasets we discovered a few started to become bloated past the point of peak functionality within a few months of even casual use.
While it's possible that it's the inhouse .NET code that could be creating the problem, after several weeks of debugging we're pretty sure it's BOINC related. My suggestion is to steer clear and look for a safer and more reliable API (or roll your own).
Try not. Do or do not, there is no try.
-- Dr. Spock, stardate 2822-3.
No. You'd have the effect before the cause would be visible. The cause, however, would still precede effect.
No, for at least some moving observers you do wind up with the effect preceeding the cause. It's all part of relativity. Two observers moving in opposit directions can dissagree about the order of two events. If anything exceeds the speed of light one of the observers will see the effect preceed the cause on the time line.
There is no such thing as "simultaneous", it's all relative.
-
- - You can't take something off the Internet! That's like trying to take pee out of a swimming pool.
As written up at the back of Wired mag a few years back.
http://www.geo600.uni-hannover.de/
Picture two tubes, each exactly 600m long and at 90 degrees to one another in the horizontal plane. Bounce a laser beam off a mirror at the end of each one. The time should be identical. Unless there is a gravitational pulse, in which case one would appear shorter than the other.
Or maybe this is something completely different =)
No, I did not read the f***ing article!
Several companies have tried to create a commercial grid software setup that pays users for their contributions. None of them have taken off. They have trouble getting customers because they don't have an existing user base waiting to crunch. (It makes your sales a little more difficult when you can't say "We can get started immediately." Instead they have to say "If you pay us money, then we'll be able to go out start trying to get all those end users to sign up.")
The end users meanwhile don't want to sign up to run endless amounts of "test packets" that aren't accomplishing anything. (They obviously don't start getting paid until there's actually customers to crunch for.) It also doesn't help that these companies' software was also kind of bloated and quirky.
The lure of being able to materially contribute to real science, in areas that are typically underfunded, by donating only idle CPU cycles is quite strong. People will do that for free. The minute you start making them focus on it as a business venture, they start getting very picky and a lot less tolerant.
I don't think you're wrong, I think there will be some pay-to-crunch type systems existing in the future. But I think they will only be branches off an existing donated network (like Seti@Home). I really doubt anyone will be able to start one from the ground up as a business model. BOINC might be a place to start, but it would need some serious modifications.
For one thing, the BOINC credit system is based on what the end users' computers self-report. Each client software runs benchmarks of its CPU, and then based on the amount of time it took to finish a Work Unit, reports back to the server how many CS (credits) it should be granted. To guard against cheating, the server will send out the same Work Unit to 3 clients, and all 3 clients will only be granted the smallest number of credits of what the 3 individuals claim.
It will probably work well most of the time, because you have millions of users, and no real incentive for most of them to cheat. The probability of the same packet being sent to 3 different cheaters is fairly small. (And even if all 3 WERE cheaters and got more credits than they deserved, it doesn't REALLY matter, does it.)
But in a commercial setup, 100% of your end users have an incentive to cheat. (If you're getting paid $1.50 per credit, it's in every end users' interest to claim as many credits as you can get away with, regardless of how long it actually took.)
But regardless, I think distributed computing projects are going to be taking off dramatically in the next few years, paid or otherwise. It's going to be pretty exciting to see the kinds of crazy things people will start wanting to crunch with it.
For a while I worked as a research programmer for one of the General Relative Groups working on the GEO600 Gravitational Wave Detector in both the UK and Germany. GEO600 is a UK and Germany co-project.
The interferometer is a typical Michaelson interferemoter using lasers with two orthogonal branches 600 metres in length. These gravitation events are small. Movements are ~10-E24 metres. It is expected that only one or two events a year will be detected. So it must run 24/7, 365 days a year.
Naturally you have to remove as much of the noise from the data as possible to detect an event. Mirrors are hung on glass threads as they are thermally inert. It runs in a vacuum. It is temperature controlled. Everything is monitored from air pressure to sisemology. The amount of data being produced is incredible. I assume LIGO is the same hence the distributed analysis.
GE0600 uses a microwave link to transmit data from the site to Hanover where it is backed up and fat pipes pass it on to partner universities. The 'head end' on site uses triple redundancy and enough bufferage for 24 hours back-up on site.
You are talking many gigabytes a day and many terabytes a year and some where in this lot will be an event. This is truely the domain of super computing or distributed processing.
Of course, even LIGO which is larger, is unlikely to spot many events if any and we will probably have to wait until LISA, the NASA/JPL/ESA spaced based interferometry project is up and running to get decent results.