Report long-running models here

Message boards : Number crunching : Report long-running models here

To post messages, you must log in.

Previous · 1 . . . 11 · 12 · 13 · 14

AuthorMessage
P . P . L .

Send message
Joined: 20 Aug 06
Posts: 581
Credit: 4,865,274
RAC: 0
Message 68184 - Posted: 24 Oct 2010, 5:01:46 UTC

This one was near enough to 7hrs on my 4hr runtime pref, on my x6 1055t.


PCS_calmodulin_v1.frag_23-147_SAVE_ALL_OUT_22378_17_0


https://boinc.bakerlab.org/rosetta/workunit.php?wuid=341596781

======================================================
DONE :: 2 starting structures 24572.3 cpu seconds
This process generated 2 decoys from 2 attempts
======================================================

p.s/ Not good credits either!


ID: 68184 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Chris Holvenstot
Avatar

Send message
Joined: 2 May 10
Posts: 220
Credit: 9,106,918
RAC: 0
Message 68202 - Posted: 26 Oct 2010, 21:02:37 UTC

I seem to be getting some "sub-optimal" results with tasks with a name in form of:

PCS_xxxx_atensor.frag....

For example - Task ID 373876225 - running on a dedicated Phenom II core clocked at about 3.0 ghz with adequate memory netted only 39 credits for a bit over 10 hours run time - CPU time, not wall clock time. Which is "watchdog territory" as my run time is set to a modest 6 hours.

I have in the past reviewed how credits are apportioned and even think I understand it and the way things are "averaged", however if there is a "non-cryogenic" CPU out there who can process Rosetta at a rate 8 to 10 faster than my Phenom II does (which is what it would take to drop my calculated average that much) I want to look into buying a few ...

ID: 68202 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
transient
Avatar

Send message
Joined: 30 Sep 06
Posts: 376
Credit: 10,733,571
RAC: 3,079
Message 68207 - Posted: 27 Oct 2010, 5:17:23 UTC

One or a few models that run long don't have a lot of influence on the credit average. These models deserve attention because they run long. Low credits are one of the consequences of this long runtime. The long runtime is not necessarily due to the type of CPU it is run on.

Maybe the CPU the WU is run on is a factor, but I don't think so. If I remember correctly, they occur with all types of CPU.
ID: 68207 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 68216 - Posted: 27 Oct 2010, 16:13:01 UTC

Right, no other CPU type would have done any better with the identical task. But the low credit you received reflects that the average tasks out there for that same line of work are not as intense as you one you happened to receive. Over time, you get your share of "easy" tasks as well, and the opportunity to earn "easy" credits (think about the the 20 tasks that preceded the one you mention, and how if each were just 3% more credit then you'd otherwise expect, it offsets the problem one... it's just impossible to see the 3%).
Rosetta Moderator: Mod.Sense
ID: 68216 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Chris Holvenstot
Avatar

Send message
Joined: 2 May 10
Posts: 220
Credit: 9,106,918
RAC: 0
Message 68223 - Posted: 27 Oct 2010, 23:32:13 UTC

Mod.sense - as usual, my reason for asking was a concern about a dysfunctional task burning CPU and not accomplishing much - I've wrote a few programs like that over the years.
ID: 68223 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Bikermatt

Send message
Joined: 12 Feb 10
Posts: 20
Credit: 10,552,445
RAC: 0
Message 68225 - Posted: 27 Oct 2010, 23:41:40 UTC
Last modified: 27 Oct 2010, 23:55:06 UTC

Chris, I have noticed that the PCS_ tasks run very slow in Linux. On my 2.2 GHz Linux box they were taking 10 hours to make two models. On my 2.1 GHz Win 7 box they always seem to make at least 4 models in 6 hours.

A few days ago I was getting a ton of them so I put my Linux machines on WCG for awhile but you can look at the results for my Win 7 box and pick out the PCS tasks just by looking at the granted credit.

Edit: In fact, I have looked at a lot of other Win 7 boxes out there and all of the PCS task on Win 7 seem to be getting much higher granted credit than what was claimed. So maybe it is some kind of dysfunction?
ID: 68225 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Chris Holvenstot
Avatar

Send message
Joined: 2 May 10
Posts: 220
Credit: 9,106,918
RAC: 0
Message 68231 - Posted: 28 Oct 2010, 11:58:08 UTC

Hey Bikerman -

Thanks for the interesting information. And here I was sitting back all smug when all the Windows guys were having all those memory problems with the miniRosetta 2.15 application - and my Linux / OS X boxes were purring along fat, dumb, and exceedingly happy.

Kind of makes you wonder if the developers spend a truckload of time looking at cross-platform issues.

Mod.sense - off the top of your head do you know which compiler (GCC??) and version they use to put the apps together and if they compile with debug symbols on? If so we can probably get a profile and see if there is a specific routine burning the cycles.

ID: 68231 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 68236 - Posted: 28 Oct 2010, 15:48:13 UTC

No, I do not know that level of detail. I am just a volunteer, not a member of the Project Team. But I would just say that everything is set up so that when they see a problem report such as yours, they can set up a WU locally and run the exact same models that your machine did, and therefore take their own observations under any debug environment that is appropriate.

This is why the problem reports are important. The feedback in the form of a fix or a new release or better credit or more consistent model runtimes is not generally immediate, but it does come.
Rosetta Moderator: Mod.Sense
ID: 68236 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile AdeB
Avatar

Send message
Joined: 12 Dec 06
Posts: 45
Credit: 4,303,881
RAC: 132
Message 68297 - Posted: 31 Oct 2010, 11:43:28 UTC - in response to Message 68225.  

Chris, I have noticed that the PCS_ tasks run very slow in Linux. On my 2.2 GHz Linux box they were taking 10 hours to make two models. On my 2.1 GHz Win 7 box they always seem to make at least 4 models in 6 hours.

A few days ago I was getting a ton of them so I put my Linux machines on WCG for awhile but you can look at the results for my Win 7 box and pick out the PCS tasks just by looking at the granted credit.

Edit: In fact, I have looked at a lot of other Win 7 boxes out there and all of the PCS task on Win 7 seem to be getting much higher granted credit than what was claimed. So maybe it is some kind of dysfunction?


Here too some "low credit" PCS_-tasks:
resultid=375503932
resultid=374981802

OS = linux
CPU = AMD Phenom II X4

AdeB
ID: 68297 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Saenger
Avatar

Send message
Joined: 19 Sep 05
Posts: 271
Credit: 695,966
RAC: 187
Message 68304 - Posted: 31 Oct 2010, 17:44:07 UTC

I've got some of those low-paying buggers as well on one computer, as on the other one other projects run as well, and the WUs are set to 20h, it will take some time yet for the one there to finish.

This one (celldivs_LL_1de2_2nn8_ProteinInterfaceDesign_26Oct2010_22394_263_0) got 0,22 credits for an hour of crunching, this one (celldivs_LPr_1de2_1v7r_ProteinInterfaceDesign_26Oct2010_22395_23_0) got 6,78, appropriate would have been ~20 per WU or hour.
And this two (PCS_2RN2_v1.frag_41-81_SAVE_ALL_OUT_22378_13_0 and PCS_CGR26A_trim_v1.frag_73-129_SAVE_ALL_OUT_22378_29_0) crunched for several hours despite the 1h limit set to the venue and got credits like it crunched just this 1h.
Grüße vom Sänger
ID: 68304 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Saenger
Avatar

Send message
Joined: 19 Sep 05
Posts: 271
Credit: 695,966
RAC: 187
Message 68361 - Posted: 3 Nov 2010, 5:36:58 UTC - in response to Message 68304.  

I've got some of those low-paying buggers as well on one computer, as on the other one other projects run as well, and the WUs are set to 20h, it will take some time yet for the one there to finish.

OK, that one was one of the faulty ones as well. It finished tonight and got far too little credit without doing something obviously wrong.
ID: 68361 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Snags

Send message
Joined: 22 Feb 07
Posts: 198
Credit: 2,367,352
RAC: 1,170
Message 68630 - Posted: 15 Nov 2010, 18:34:32 UTC

1NKU_R2_LESSPCSCST2_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22545_14134_0

3 models completed in 56635.9 cpu seconds (cpu_run_time_pref: 43200)

Looks like the first two models would have taken 4 hours or less each while the third took 7+ hours.


Best,
Snags
ID: 68630 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,155,769
RAC: 1,645
Message 68631 - Posted: 15 Nov 2010, 18:54:48 UTC

Me too,

AR3436A_blind_LESSPCSCST_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22535_647_0

First model of this task has already taken nearly 12 hours of CPU and is still not complete. On step 177,000 with 24hr runtime preference.
Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 68631 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 . . . 11 · 12 · 13 · 14

Message boards : Number crunching : Report long-running models here



©2021 University of Washington
https://www.bakerlab.org