A beowulf cluster with nothing to do.

September 12, 2009 3:06 PM

I scrapped all my junk computers together and made the most rag tag beowulf cluster ever in my living room. Mostly as a conversation starter. But now I have to figure out what to have it working on. What's something that will sound really cool?
posted by brenton to computers & internet (10 comments total)

Just do folding@home and say that you're helping to calculate the cure for cancer.
posted by sonic meat machine at 3:13 PM on September 12 [1 favorite]

Seconding folding@home. You might find something flashier, but it'd be flashy at the expense of helping cure cancer.
posted by Tomorrowful at 3:22 PM on September 12

You really need to do a watts/MIPS calculation here.
posted by Leon at 4:16 PM on September 12 [4 favorites]

Sure, curing cancer is cool. But perhaps finding aliens is cooler?
posted by jeoc at 4:16 PM on September 12

How about trying to brute-force crack the remaining unknown keys for AACS? (You'd have to write your own code, though.)
posted by Chocolate Pickle at 4:31 PM on September 12

Hard to beat a cure for cancer, but how about the Great Internet Mersenne Prime Search?
posted by whiskeyspider at 4:58 PM on September 12

Hm, unfortunately our cluster is written in a special way so we can't just hook an existing program into it because it wouldn't work. We have to write it from scratch. So no cancer research. :(
posted by brenton at 11:26 PM on September 13

For writing from scratch, is video editing / generation up your alley? because it is pretty easy to use the ffmpeg library to make a small app that reads a video file, and once you have a few programs reading and processing individual videos, you can start playing with combining video feeds. The gstreamer plugin architecture may also be worth looking into, because not only will it provide an interface for reading media, but also for mixing, processing, and sending over a network.

Another possibility that can readily take advantage of multiple physical machines / CPU's is machine vision. For example with two cameras set up properly you can get 3d input, and there are an infinite number of cool things you can do with 3d spacial correlation. (especially if you mix this with MIDI / servos / x10 / etc. some kind of physical output ). You could split up parallel searches for visual features across the multiple CPU's quire readily.

Or, even better, you could combine spacial recognition and video synthesis and produce a program that tries to recognize shapes and then resynthesize its input by finding video frames in its media library that can be rotated or resized to have similar shapes - so it would collage together images in real time of whatever you put in front of the camera made of its library of film footage. There are a few open source shape recognition libraries out there.

But really, do whatever you think is cool, because if you are programming it yourself, you will need it to be really fucking awesome to have the motivation to put in the amount of work it will take to finish it.
posted by idiopath at 5:49 PM on September 21

idiopath: thanks for the suggestion. sounds awesome. perhaps a bit beyond my abilities at this point, but good thoughts to keep in mind for the future.

for now, we're calculating primes using the Lucas method, then looking at the results statistically to try to find a way to increase the probability of finding a prime. so we're "calculating big prime numbers" which sounds about as cool to me as folding@home, though admittedly less altruistic.

leon: (if you're still following this thread) what would a watts/MIPS calculation tell me? how would I do such a calculation?
posted by brenton at 9:22 PM on September 30

Brenton: sorry, I was too terse.

I was suggesting that new hardware might use less electricity to do the same amount of work than old hardware - essentially, that the modern hardware will pay for itself in reduced electricity costs.

To do the calculation... I guess I'd plug the PC into a watt meter, have it process n blocks of data, then run the same test on another PC. That should tell you which PC processes your data most cheaply, and what the savings will be over a given time period.
posted by Leon at 2:10 PM on October 4