[vault_]

I've done this at my university, but in a slightly less obnoxious way.

We have a grid computing environment set up (using Sun Grid Engine) that allows people to run bulk jobs across both dedicated clusters and in spare cycles on lab machines. The way it was set up, CPU jobs on lab machines were set to an extremely low priority, so they don't interrupt legitimate use, and the jobs in some queues were set up so they'd be suspended when real jobs needed to run.

I'd submit hundreds of jobs to the background queue, which would run for a couple of hours and then stop. I also had access to a fairly large number of Nvidia GPUs which were used largely to teach people Cuda and run big simulations every month or so. I was able to use a pretty good portion of my college's compute power without being particularly annoying to other users.

I stopped after a while (maybe 2 weeks), because it was rather inefficient (I mined about 0.04 coins, at the time worth about $8) and because it was a huge pain to maintain. I would probably do it again with a different currency, but it's really not worth it.