I've just received a polite email from the manager of the computer cluster I'm using, asking that I learn to run my jobs properly (by submitting them to the queue) rather than just running them from the terminal. The only reason I haven't been doing the right thing is that I don't know how.
So he pointed me to a web page with instructions, which were rather opaque to me because I don't really speak Unix. So then I asked the post-doc who's been doing this with her Perl program, and she pointed me to a set of instructions I'd scribbled down about 18 months ago, when I first learned about this computer cluster from a whiz-bang grad student in another lab.
So I've more-or-less blindly followed these instructions, and submitted my first job to the queue. I know I did that part correctly because I can ask the terminal to show me the queue and my job is indeed in it, along with about a zillion other jobs. Mine is 53rd, so there are a zillion-53 jobs after mine. There are also about half a zillion jobs already running, so maybe being 53rd is pretty good. Some of the jobs that are running aren't predicted to finish for more than a week.... (You can probably tell that I'm out of my depth here.)
The really annoying thing is that I don't get any feedback until my job is done, when I'll get an email. Guess I'd better learn how to work this system.
Lessons on management styles from Edward Teller, Hans Bethe and Robert Oppenheimer: A question of temperament
2 days ago in The Curious Wavefunction