climateprediction.net home page
Problem running ONE SPECIFIC project
Problem running ONE SPECIFIC project
log in

Advanced search

Questions and Answers : Windows : Problem running ONE SPECIFIC project

Author Message
nzmeccanoman
Send message
Joined: 22 Oct 10
Posts: 1
Credit: 458,816
RAC: 0
Message 54759 - Posted: 7 Sep 2016, 3:47:32 UTC

I have many projects running on 2 computers that behave as expected ie for every hour elapsed the time remaining drops by roughly an hour) but one of 4 climate prediction work units is behaving very strangely. Is there an issue with this project? Should I abort it. It worked well until it got to roughly 55% completed with about 50 hours to go Since then for every hour run time the time remaining increases indicating that the work unit is running far slower. Currently on 64 % complete.(but still increasing) Elapsed time now 272 hours 27 minutes. Estimated remaining time now 62 hours 30 minutes. All other Climate prediction and work units from 4 other groups, as this did till half way through cheers
____________

Les Bayliss
Volunteer moderator
Send message
Joined: 5 Sep 04
Posts: 6408
Credit: 16,839,542
RAC: 21,887
Message 54760 - Posted: 7 Sep 2016, 4:14:38 UTC - in response to Message 54759.

Could you provide a link to the model page of the task in question please.

Profile Iain Inglis
Volunteer moderator
Send message
Joined: 16 Jan 10
Posts: 877
Credit: 100,083
RAC: 3,242
Message 54762 - Posted: 7 Sep 2016, 19:18:11 UTC

You've got one Mexico (25 km resolution): if you have access to that PC could you report back the "gflops" for that model? To do that just look at the model's properties in BOINC Manager.

The reason for asking is to compare gflops estimates supplied by the project with actual run-times for all model types - and I haven't had a MEX25 so don't know how it's sized.

Thanks.

Profile Alan K
Send message
Joined: 22 Feb 06
Posts: 203
Credit: 10,727,650
RAC: 8,191
Message 54763 - Posted: 7 Sep 2016, 22:04:47 UTC - in response to Message 54762.
Last modified: 7 Sep 2016, 22:05:26 UTC

If it's any help one of the mex25's that I have run took 821,000 (approx) secs cpu time on a 3.5GHz i5. Approx 5.4 sec/time step.

Profile Byron Leigh Hatch @ team Carl Sagan
Send message
Joined: 17 Aug 04
Posts: 270
Credit: 38,039,661
RAC: 64,767
Message 54764 - Posted: 7 Sep 2016, 22:07:53 UTC
Last modified: 7 Sep 2016, 22:41:21 UTC

_


also If it's any help ... my one and only mex25 still running:
my mex25 should finished in 2 days 13 hours,
on an Intel Xeon CPU E5-2687W v3 @ 3.10GHz

wah2_mex25_c21x_201112_13_410_010611137_1

Screen shot of properties:



wah2_mex25_c21x_201112_13_410_010611137_1

I hope this will help in some way.

_

Profile Iain Inglis
Volunteer moderator
Send message
Joined: 16 Jan 10
Posts: 877
Credit: 100,083
RAC: 3,242
Message 54765 - Posted: 7 Sep 2016, 23:31:00 UTC
Last modified: 7 Sep 2016, 23:35:30 UTC

Exactly what I needed, Byron. Thanks.

I don't want to trigger a discussion about differential credit rates - aargh! - but systematically describing the run-times of different tasks is of general interest. The HADAM3P AFR tasks were wrongly sized, if I remember rightly, and caused a number of comments such as nzmeccanoman's (though I failed to record the gflops for AFR50 despite having run loads).

The only WAH2 task I don't have a gflops value for, but which isn't very common, is the PNW/25-km/16-month.

bernard_ivo
Send message
Joined: 18 Jul 13
Posts: 252
Credit: 5,903,045
RAC: 23
Message 54767 - Posted: 8 Sep 2016, 5:30:07 UTC - in response to Message 54765.
Last modified: 8 Sep 2016, 5:39:21 UTC

I don't want to trigger a discussion about differential credit rates - aargh! - but systematically describing the run-times of different tasks is of general interest.

The only WAH2 task I don't have a gflops value for, but which isn't very common, is the PNW/25-km/16-month.


Hi Iain,
here are the properties of one wah2_pnw25_a3ru_20069_16_424_010649978_1, running on WIN 7, i5-2520M @ 2.5 GHz, Average (sec/TS) = 2.20

Edit: Seems I'm not sure how to display the image here is the link

I used to cross-reference Average (sec/TS), CPU run times, CPU models and CPDN models in order to determine best CPU value for money. I was using excel, but I drop it at certain point as too time consuming.

Profile Iain Inglis
Volunteer moderator
Send message
Joined: 16 Jan 10
Posts: 877
Credit: 100,083
RAC: 3,242
Message 54768 - Posted: 8 Sep 2016, 8:34:36 UTC - in response to Message 54767.

Thanks, Bernard. That completes the set of WAH2 models - at least the ones I know about (i.e. DUB25/13, EU25/12-13, MEX25/13, MEX50/12-13, PNW25/16-24, SAM50/13 and SAS50/13).

WB8ILI
Send message
Joined: 1 Sep 04
Posts: 98
Credit: 40,088,982
RAC: 54,475
Message 54769 - Posted: 8 Sep 2016, 12:14:50 UTC

nzmeccanoman -

It has been my experience that if the percentage complete is increasing you are probably OK. If the percentage complete is "stalled" for a long (whatever that is) period of time, consider shutting down the BOINC Manager (stop running tasks) and restarting.

Also, the time to completion does not always match the elapsed time and percentage complete. For example Elasped=400, %complete=50, Time Remaining=100.




____________

Questions and Answers : Windows : Problem running ONE SPECIFIC project


Main page · Your account · Message boards


Copyright © 2017 climateprediction.net