I thought we were in control of run times on this project? So why is it with these tasks with my run time set at 8hrs the tasks theorize 11hrs to completion and I see one task that is at 9:08 and still has 3:21 to go?

It would appear some of these tasks are ignore the time deadline I set. Others when I checked the results were terminated before the 8hrs were up as would be expected if the next cycle of the task would exceed the time limit. The task that is running over is rb_04_13_30577_61877 etc with ignore_the_rest_47273_4336_0.

Time now is 9hrs 13 mins and 3hrs 15 minutes to end. This is far beyond the 8hrs I set.

I thought we were in control of run times on this project? So why is it with these tasks with my run time set at 8hrs the tasks theorize 11hrs to completion and I see one task that is at 9:08 and still has 3:21 to go?

It would appear some of these tasks are ignore the time deadline I set. Others when I checked the results were terminated before the 8hrs were up as would be expected if the next cycle of the task would exceed the time limit. The task that is running over is rb_04_13_30577_61877 etc with ignore_the_rest_47273_4336_0.

Time now is 9hrs 13 mins and 3hrs 15 minutes to end. This is far beyond the 8hrs I set.

Think of it more as a request than a deadline - Rosetta can't know how long a task will take until it's been completed, and if the first 10 decoys take 30 mins each, then Rosetta will rightly calculate that it should run another, which might then take four hours, taking you over the limit.

Having said that, out of the fifty or so tasks your computer is showing as complete I can only see one task that has gone over 8hrs (28,800s) and that went over by 85s, so your current one must be an odd-ball.

The 'time to complete' is calculated by BOINC and basically doesn't work properly with Rosetta.

Danny

---

I thought we were in control of run times on this project? So why is it with these tasks with my run time set at 8hrs the tasks theorize 11hrs to completion and I see one task that is at 9:08 and still has 3:21 to go?

It would appear some of these tasks are ignore the time deadline I set. Others when I checked the results were terminated before the 8hrs were up as would be expected if the next cycle of the task would exceed the time limit. The task that is running over is rb_04_13_30577_61877 etc with ignore_the_rest_47273_4336_0.

Time now is 9hrs 13 mins and 3hrs 15 minutes to end. This is far beyond the 8hrs I set.

Think of it more as a request than a deadline - Rosetta can't know how long a task will take until it's been completed, and if the first 10 decoys take 30 mins each, then Rosetta will rightly calculate that it should run another, which might then take four hours, taking you over the limit.

Having said that, out of the fifty or so tasks your computer is showing as complete I can only see one task that has gone over 8hrs (28,800s) and that went over by 85s, so your current one must be an odd-ball.

The 'time to complete' is calculated by BOINC and basically doesn't work properly with Rosetta.

Danny

ok..a whopping 85 seconds i can live with.
Just the completion times are really out there.
I don't see anything abnormal either. Weird how something showing a run time of 9hrs+ comes up with an end result of just a tiny bit over 8hrs. oh well.
Thanks for the answer

Could it be the difference between CPU time and wall clock time?

wall clock? not sure what you mean.

I mean the time the actual computation takes as opposed to the difference between the time the task starts and the time the task stops. There could, for example, be a period that the app waits for enough memory to be available before it can continue actual computation.

ok but what I referred to was actual run time clock.
if the program sits still then the run time clock sits still correct?
the actual run time clock showed 9hrs but then when it reported the time used was only 8hrs.

You can look at the specific task's properties and see the total runtime (elapsed time, wall-clock time), and the actual CPU time.

BOINC's estimates for time to completion are scaled to what it has been seeing on prior tasks. If you have one task run long (Rosetta's watchdog permits a task running up to 4hrs longer than your runtime target. This would only occur if the last model processed takes significantly longer than the average per model times of the first models crunched on the work unit.) then BOINC sees, for example, an 11hr runtime, and starts to bend it's scaling towards 11hrs.

If you have an 8hr runtime preference set, it takes BOINC a few tasks to bend it's scaling from the default 3hrs it probably took for the first few tasks from Rosetta, up to the new 8hr observations. Similarly, assuming your next several tasks run in the more typical 8hr timeframe, BOINC's estimates will be scaled back from the present 11hrs back down near 8.

The Rosetta watchdog assures your preference is never ignored. But it is not intended to be a hard limit either. Most of the time, the limit is honored without any problem. For some specific types of work units (especially for new protocols as they are being developed), specific models can take many times longer than the normal. The watchdog assures that these do not run endlessly. If that last model completes, another will not be started. So for a task to run for 4 hrs beyond your preference, it has spent 4hrs or more on a single model. This is significantly longer than a single model generally takes for any of the Rosetta protocols, and the point at which the watchdog will step in and end the task.

---

Rosetta Moderator: Mod.Sense

Thanks Mod.Sense
That clears things up for me.