A fun idea came to me yesterday, and it's easy to formulate. Right now there are projects where people provide their computers for calculations. You install a client program, it runs all the time and, analyzing the sent data, it does something. Something useful, like searching for a cancer cure, or of little use, like searching for aliens. You could also do some genome decoding, this will bring closer the coveted era of universal genetic modification. All these projects are partially based on server technology, rather than p2p, because the server distributes and collects the tasks. But why not make a strictly p2p project, where participants in a grid provide services to each other, not by sharing Britney Spears porn, but by providing CPU time for other people's tasks?
The easiest way is to apply this idea to CPU-intensive games like chess (I don't know, truth be told, how easy they are to parallelize). I mean, people won't be playing all that much, there will be enough CPU time for everyone if the client program is on all the time, and there are many participants in the grid.
The sad part: of course, all this has already been invented before me. Here, for example.
p2p computing for games
2007-05-08