|
What is APS@Home? What does it do?
Is APS@Home related to Climateprediction.net (CPDN) or the BBC climate experiment?
What's this "BOINC" thing I see mentioned on the site?
APS@Home is a volunteer desktop grid computer system which solves problems in atmospheric science. At the moment, we're concentrating on atmospheric dispersion, and how it affects the accuracy of estimates used in global climate models. In the current application, your computer simulates the trajectories of tracer "particles" within and above forest and crop canopies, to give us some understanding of the surfaces which measurements of ecosystem - atmosphere interactions are made above. These measurements are eventually used as input to climate models, and the understanding we gain from the work your computers do helps to improve the quality of these input data. There are plans to diversify into cloud microphysics and other areas eventually.
The APS@Home server sends software to your computer, along with a few input files which describe a problem we'd like to solve. When your computer isn't busy (and you get to decide what that means), the software runs on your unused CPU. This all works using the same software as SETI, the BBC Climate Change Experiment, and many others. It's a well tested system which has run for many years without problems. The only difference is the types of problems we try to solve. See the "Science" message board for more information.
Credit is a way of measuring your contribution to the project. The more work your computer(s) do, the more credit you get. Some people like to compete amongst themselves to see who can make the largest contributions to certain projects. There's a worldwide network of teams, some of whom take part in APS@Home. You can read about them in the "Team Talk" section of the message boards.
In the current application, your computer simulates the trajectories of particles within and above forest and crop canopies. The application keeps track of the number of trajectories you calculate, and how long each one is. We then award credit based on the total number of time steps your computer has completed over all trajectories. It takes just over half a million floating point operations to complete a single time step. There are typically 10,000 to 100,000 steps per trajectory, and several hundred trajectories per work unit. That's why you'll see some variability in the amount of credit awarded, - the length of trajectories is almost random.
We don't use benchmarks to calculate credit. This is because some volunteers use optimised clients, whose benchmark results don't accurately reflect the speed at which the host runs the APS@Home applications. This means that volunteers with optimised clients get more credit for the work they do than other people. This isn't fair, and doesn't reflect the contibutions made by different people to the APS@Home goals.
There are sometimes valid reasons for using optimised clients, - the main one being where an optimised science application is provided. In this case, optimised clients give a more accurate estimate of the speed at which an optimised application runs. However, we don't release applications optimised for specific architectures. So that people can use the same BOINC client to contribute to other projects as well as APS@Home, we decided to avoid this whole minefield. The credit you are awarded reflects the amount of APS@Home science you have contributed, regardless of optimisations made for other projects.
APS@Home is not affiliated or related to CPDN or the BBC climate experiment, although we're great admirers of their work. Both of those projects run stripped down global climate models, which is why run times were sometimes so huge. We focus on very small scale, short duration processes in the atmosphere, which is why our run times are shorter. You may decide to contribute to other projects as well as APS@Home, to only run APS@Home, or to only concentrate on other projects. Our goals are broadly similar, and whichever projects you choose to contribute to, you'll be helping us to understand the Earth - climate system.
BOINC is an acronym for the Berkeley Open Infrastructure for Network Computing. It's a set of software and tools which let our server talk to your computer, to distribute applications, award credit, and return results. It's used by many distributed computing projects including CPDN and even SETI. The BOINC homepage is here.