photo via Richard Bell
MilkyWay@home is a distributed computing project, harnessing the power of volunteered computers to create a highly accurate 3D model of the Milky Way galaxy. The project uses data gathered by the Sloan Digital Sky Survey (SDSS) and the Berkeley Open Infrastructure for Network Computing (BOINC) platform. By volunteering a percentage of your computer’s unused operating power, your computer will focus on mapping out a small section of our galaxy. MilkyWay@Home’s data, source code, research and results are open source and available for public use.
In general, an astrophysics problem revolves around creating a computer system model that will replicate what we see in the sky – if a model matches exactly then we can leapfrog off that information to work on a bigger, more involved problem. MilkyWay@Home is currently focusing on the Sagittarius stream, which provides knowledge about how our galaxy was formed and how tidal tails are created when galaxies merge. The general idea is that our galaxy actually has smaller galaxies mixed within it, probably from galactic collisions in ancient history. Mapping the dynamics of such interstellar streams is expected to provide crucial clues for understanding the evolution of the Milky Way and similar galaxies. It could also provide insight on dark matter.
This project also enables research in both astroinformatics and computer science. Astroinformatics is a rising field at the interface between Computer Science and Astronomy with new discoveries made possible by the abundance in galactic data created by the SDSS. As of January 2010, MilkyWay@Home has 44,900 users and 1,590 teams in 170 countries, with an average computing power of 1,382 TeraFlops. This ranks MilkyWay@Home second amongst the Top 500 list of supercomputers.
Project owners + coordinators:
Rensselaer Polytechnic Institute’s departments of Computer Science and Physics/Astronomy
Travis Desell, Graduate Research Assistant in Computer Science
Dave Przybylo, Undergraduate Research Assistant in Computer Science
Nathan Cole, Graduate Research Assistant in Physics, Applied Physics and Astronomy
Malik Magdon-Ismail, Associate Professor of Computer Science
Heidi Newberg, Associate Professor of Physics, Applied Physics, and Astronomy
Boleslaw Szymanski, Claire and Roland Schmitt Distinguished Professor of Computer Science
Carlos Varela, Associate Professor of Computer Science
posted on Aug 13, 2010:
I am deactivating this project because each work unit takes WAY too much time to complete. When transfered down, each unit is typically 135 hours on my iMac – that’s over 4 and a half DAYS, folks… along with a pointlessly accelerated due date not much longer than that, which forces it into immediate ‘emergency’ run mode and hogs my machine. If an error occurs, you lose ALL that time – totally unacceptable. Hey, not everyone has a super computer on their desk. NO work unit should EVER take more than half a day to complete. If ‘management’ would slim these units down (by a factor of 10!), I would resume my participation. THANKS, Darryl
posted on Aug 17, 2010:
I think they should withdraw their CPU app.
It is very slow and they don’t need it.
A CPU task from this project takes 16-18 hours on a Pentium 4 2.40GHZ.
The same task takes 4 minutes in an ATi GPU
posted on Feb 02, 2011:
Is source code of the server part available?
posted on Feb 02, 2011:
Hi Everyone, Travis here from MilkyWay@Home. Would just like to make a few clarifications and would like to thank everyone for the support of our project. The CPU workunits ARE very useful to us, around 40-50% of all the work done on MilkyWay@Home is by CPUs. We also have a new application for using N-Body simulations to model the formation of different structures in the Milky Way Galaxy, which has rather variable compute time so these may work better on older machines.
Our server code is freely available online here
https://github.com/travisdesell/fgdo_java and here https://github.com/travisdesell/milkyway_server if anyone would like to play around with it. We’ve re-implemented the BOINC assimilator/validator in Java (see the FGDO package), so it may be of some interest. The rest of the server code is available on BOINCs webpage.
posted on Apr 26, 2011:
hi Travis,
Is anyone doing this in C++ instead of Java?
thanks for sharing the code, otherwise i will try to adapt it.
thanks
Anne
posted on May 01, 2011:
Hi Anne,
BOINC’s server code is in c++, so you could just use that. Ours is just in Java to make the evolutionary algorithms we do easier to implement.
–Travis