View Full Version : Certain Tasks are taking forever?
I just checked my Rosetta tasks and there are 2 "Proteininterfacedesign" tasks that are running for almost 8 hours now - funny thing is they claim to have only 10 minutes left, but that is steady for the last hour or so and they are progressing verrrrry slowly. Is this normal, and will they be rewarded with higher credit or should I kill them? My machine usually cranks through a unit in about 4 hours.
I leave them alone as long as they make a small amount of progress every four hours - even if it's just a VERY small amount.
What can happen is either:
1) The protein has been folding along the paths that were expected more or less, but now the simulation has found new folding patterns for the protein, that need to be analyzed.
2) sometimes the simulation just runs off the rails - an error has occurred and the wu needs to be stopped eventually.
There is the chance also, that you may have just received a larger (longer) wu. Maybe one designed for 12 hours say. That behavior of staying at the same or nearly the same completion percentage, is entirely normal for a big wu. They take more even increments in the smaller wu's, simply because each "frame" takes less time, and the percentage shown to you, is only a guess anyway, until the next frame or "step", is reached. That's why sometimes you'll actually see it move to a higher percentage figure, instead of a lesser one.
As long as they're showing high cpu usage in Task Manager, and making SOME progress every 4 hours, let 'em go.
I never worried a lot about how much I earned in points for a work unit, but I believe points are still awarded based on the time spent crunching by the faster of the two rigs who work on that wu.
One pc will crunch to establish an answer, and the second pc will crunch that same wu, to verify the answer, and compare the time spent crunching. That assures correct data, and stops any "I want points", funny business, from unethical crunchers.
Thank you for taking the time to explain this behavior - I appreciate it
You're welcome. There's a more thorough (and probably up to date) explanation, somewhere on the Baker Labs site, but I don't have the url for it.
vBulletin® v3.8.7, Copyright ©2000-2013, vBulletin Solutions, Inc.