Distributed Computing "Advances"
Quirk writes "NewScientist is reporting on..."Software to be launched in January will let PC users run as many "distributed computing" projects as they like. The program will let PC users search for aliens, help predict climate change and perform advanced biological research - all at the same time."'It is called the Berkeley Open Infrastructure for Network Computing (BOINC). BOINC acts like a software platform that can run a number of screen-saver style applications on top of the PC's own operating system.'"
So the whole work has to be done twice for the sake of correctness. I think they should introduce some trusted user mode, let's say, so that results from users who have invested a certain amount of cpu time should be trusted or at least not every received result double checked. Just every n'th packet or so and if it's invalid they have to recheck all unchecked packets. I guess this would reduce double work a lot as there is normaly only a minority of users who's trying to cheat.
Does this sound sane?
I've always had some mild reservations about running the closed-source SETI code, but convinced myself it wasn't an unreasonable exposure. A meta-app that exists to download yet more closed-source code without telling me... nope, that's over the line. Sorry, lil' green guy, but this is too much to ask.
(signed) a top 1% setiathome client.
I'd like to see a distributed computing app that can be used to both do the work (like the current ones do), AND optionally have the ability to submit a task. This way you could have a world wide supercomputer that everybody would have a chance to employ. Very few people would probably use it, but it would be very interesting to see the ways in which different people put it to use.
I've been thinking about something like this all semester in my Distributed Computing class.
:)
What I'd really like to see is a system setup where you have a network of clients, any of whom can dispatch an agent across the system that consumes resources to accomplish some goal.
Obviously there would have to be some sort of non-malicious code signing or sandboxing going on within the system, as well as forcing the agents to consume proportional resources (ie the more time/space/bandwidth you give to the sytem, the more you can consume)... either way it's still a neat idea that I'd be eager to participate in...
It'd be a little more exciting that Folding at anyrate..
My Folding@Home Team