Message boards : Number crunching : huge WU
Author | Message |
---|---|
hulk189 Send message Joined: 14 Oct 05 Posts: 2 Credit: 28,789 RAC: 0 |
what's with the huge work units normally a I can go through a work unit in 3-4 hours now I have one showing 30+ hours. |
David Baker Volunteer moderator Project administrator Project developer Project scientist Send message Joined: 17 Sep 05 Posts: 705 Credit: 559,847 RAC: 0 |
what's with the huge work units normally a I can go through a work unit in 3-4 hours now I have one showing 30+ hours. that is no good. what is the name of the WU? |
hulk189 Send message Joined: 14 Oct 05 Posts: 2 Credit: 28,789 RAC: 0 |
production_abinito_1ten_250_1979_0 it's now down to about 13 hours. |
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
production_abinito_1ten_250_1979_0 it's now down to about 13 hours. When a WU first arrives on your system, depending on a number of factors local to your machine, BOINC will provide an estimate of the time to complete the WU. As processing proceeds, this estimate will adjust to a more accurate estimate. The adjustments occur at WU check points, which you can identify as the points at which the percent compete moves up. When the WU first starts the percent complete will move to 1% right off. It will them be quite some time before it moves again.... Usually this second move will be a jump to 10%. Between jumps you will actually see the time to completion RISING, only to fall back at the next jump in percent complete. If your work units are arriving with a estimated time of 30 hours (some will some won't) then one of the numbers the system uses in the estimate calculation is too high. It will adjust itself over the course of a number of WUs. The problem is made worse if you process a number of short WUs followed by a long one or vise-versa. This throws the adjustments out of whack for the next few WUs. You should be ok unless the WU fails to compete with a "Maximum Time Exceeded" error. This can happen but adjustments have been made to prevent this kind of WU failure. If you have more problems with it, be certain to post about it on this thread so the project guys can find the info. Moderator9 ROSETTA@home FAQ Moderator Contact |
anders n Send message Joined: 19 Sep 05 Posts: 403 Credit: 537,991 RAC: 0 |
production_abinito_1ten_250_1979_0 it's now down to about 13 hours. This sort of Wu-s take alot more time than others. Check this one out. https://boinc.bakerlab.org/rosetta/workunit.php?wuid=5701993 The good thing is that I have not seen any errors in that name series. Happy Crunching Anders n |
Hoelder1in Send message Joined: 30 Sep 05 Posts: 169 Credit: 3,915,947 RAC: 0 |
See this thread: nstruct=40. It seems the production_abinitio... WUs perform 40 instead of the usual 10 trajectories. Presumably, since they only perform the 'ab inito' part of the code, the WUs would have been extremely short, just doing the usual 10 trajectories -- hence the increase to 40 (just guessing). |
Team TMR Send message Joined: 2 Nov 05 Posts: 21 Credit: 1,583,679 RAC: 0 |
I know these ABINITIO WUs are big and take longer, but is anyone else finding they disappear when they're reported? We've just uploaded two of these WUs, each took 18 hours to complete, but both are missing from our list of results. I've checked back, and of the 5 ABINITIO WUs that we've completed this morning only 1 is present in our results list. Every other type of WU seems to appear in the results list immediately. I'm fairly sure we're getting credit (our credit increased by the expected amount the last time we reported out of these WUs), I'm just curious why they're missing from the reports list. |
Hoelder1in Send message Joined: 30 Sep 05 Posts: 169 Credit: 3,915,947 RAC: 0 |
I know these ABINITIO WUs are big and take longer, but is anyone else finding they disappear when they're reported? They don't disappear, but don't expect them to appear at the top of your results page. Ther Result IDs of these WUs are in the 7,200,000 - 7,400,000 range (at least in my case), so you should find them on some of the following pages (the results are sorted by decreasing Result ID). |
Team TMR Send message Joined: 2 Nov 05 Posts: 21 Credit: 1,583,679 RAC: 0 |
Thanks! I obviously didn't click through enough pages. |
[B@H] Ray Send message Joined: 20 Sep 05 Posts: 118 Credit: 100,251 RAC: 0 |
I have one of those running on the other system, now at 7:18:46 and 12.50% estimated time left by BOINC is 9:27:45. Thought I had gone longer between turning unuts in, that one will do it. WU name production_abinito_2acy-250_2206_0 And another waiting in the que production_abinito_1who_250_2203_0 But they are good for the credit as long as they finsih correct. Ray Pizza@Home Rays Place Rays place Forums |
[B@H] Ray Send message Joined: 20 Sep 05 Posts: 118 Credit: 100,251 RAC: 0 |
I have one of those running on the other system, now at 7:18:46 and 12.50% estimated time left by BOINC is 9:27:45. CPDN took a turn on the CPU, now this WU, production_abinito_2acy-250_2206_0 is up to 10:26:05 and still at 12.5%, almost tempted to kill it. If the percent even went up to 15% I would know that it is going. Ray Pizza@Home Rays Place Rays place Forums |
MikeMarsUK Send message Joined: 15 Jan 06 Posts: 121 Credit: 2,637,872 RAC: 0 |
|
Message boards :
Number crunching :
huge WU
©2024 University of Washington
https://www.bakerlab.org