Message boards : ATLAS application : Change in Credit?
Message board moderation

To post messages, you must log in.

AuthorMessage
keputnam

Send message
Joined: 27 Sep 04
Posts: 94
Credit: 4,033,925
RAC: 7,546
Message 41988 - Posted: 23 Mar 2020, 20:51:27 UTC

In the last couple of days credit seems to be dramatically lower than previously

Jobs were averaging 450 - 550 credits each

Now for the same approximate run times they are averaging 125 - 150 each

ID: 41988 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1992
Credit: 143,858,252
RAC: 96,901
Message 41990 - Posted: 24 Mar 2020, 13:51:51 UTC - in response to Message 41988.  

It looks like you changed the preferences - max #cores in this case.

BOINC uses a couple of values to calculate credit but also estimated remaining runtime.
If 1 or more of those values change it takes a while until the server finds a new balance.
The greater the change the longer it takes.
ID: 41990 · Report as offensive     Reply Quote
keputnam

Send message
Joined: 27 Sep 04
Posts: 94
Credit: 4,033,925
RAC: 7,546
Message 41992 - Posted: 24 Mar 2020, 19:45:38 UTC - in response to Message 41990.  

Really?

APP_CONFIG was last changed 2020/02/20

And <avg_ncpus>2.000000</avg_ncpus for Atlas has been 2 since I started running it years ago

This vastly reduced credit started 2020/03/21

ID: 41992 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1992
Credit: 143,858,252
RAC: 96,901
Message 41993 - Posted: 24 Mar 2020, 20:27:47 UTC - in response to Message 41992.  

I did not mention app_config.xml, did I?

See your task logs:
https://lhcathome.cern.ch/lhcathome/result.php?resultid=268299314
Device peak FLOPS 34.13 GFLOPS

https://lhcathome.cern.ch/lhcathome/result.php?resultid=268841502
Device peak FLOPS 7.98 GFLOPS

This is usually caused by a change on this page:
https://lhcathome.cern.ch/lhcathome/prefs.php?subset=project
ID: 41993 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 585
Credit: 33,205,523
RAC: 18,462
Message 41996 - Posted: 24 Mar 2020, 20:58:37 UTC - in response to Message 41993.  

I have experienced similar situation myself. The device peak flops for Atlas tasks were changed to a lot lower (from 18.17 => 5.34 or so) the memory consumption that is reported to Boinc went also from 6600 MB to 3300 MB when I changed the selection of sub-projects from Atlas+Theory to include only Atlas (nothing else was changed, the max# of CPUs remained at 4). When I changed the selection to Atlas+sixtrack everything returned to normal on next downloaded Atlas tasks. I have set the CPU count to 4 on web preferences but using app_config to run the tasks with 1 CPU, this setting was not changed when the incident happened.
ID: 41996 · Report as offensive     Reply Quote
keputnam

Send message
Joined: 27 Sep 04
Posts: 94
Credit: 4,033,925
RAC: 7,546
Message 41997 - Posted: 24 Mar 2020, 22:26:37 UTC - in response to Message 41993.  
Last modified: 24 Mar 2020, 22:28:42 UTC

I don't remember changing anything there, certainly not maxcpus

I DO use an APP_CONFIG to control that, which is why I mentioned it before and I've restarted BOINC several times recently (for service application) So the APP_CONFIG would over-ride the options there, wouldn't it?


in any case, how can that possibly affect the credit awarded for an equivalent CPU time? Doesn't CPU time account for different numbers of "processors" as oppose to wall-clock time?
ID: 41997 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 1557
Credit: 57,665,912
RAC: 203,128
Message 41998 - Posted: 24 Mar 2020, 23:59:29 UTC

Atlas have a constant GFlops of 51-52 for all Atlas-Tasks in -native atm.
The Total is:
Total average computing: 100,477 GigaFLOPS (including sixtrack,Theory and CMS).
shown on the Application-page:
https://lhcathome.cern.ch/lhcathome/apps.php
See also a dropping since a few days from 50k Credit to atm 46.5k - duration.
Atlas have a constant flow of tasks for me.
ID: 41998 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 585
Credit: 33,205,523
RAC: 18,462
Message 41999 - Posted: 25 Mar 2020, 6:36:24 UTC

I just tested this again by deselecting sixtrack from my task selection leaving only Atlas and sixtracktest active. All new Atlas tasks that downloaded have the Device peak GFLOPS 4.54 as prior to that they had 18.17.
ID: 41999 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1992
Credit: 143,858,252
RAC: 96,901
Message 42000 - Posted: 25 Mar 2020, 7:30:45 UTC

Asked David Cameron to check the job templates and if all project servers are using the same template.
ID: 42000 · Report as offensive     Reply Quote
keputnam

Send message
Joined: 27 Sep 04
Posts: 94
Credit: 4,033,925
RAC: 7,546
Message 42001 - Posted: 27 Mar 2020, 8:45:34 UTC - in response to Message 41988.  

Aaaaaaaaaand

Credit is back where it was

My last 5 WUs have all been in the 450-600 credit range


<shrug>
ID: 42001 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 585
Credit: 33,205,523
RAC: 18,462
Message 42002 - Posted: 27 Mar 2020, 12:33:07 UTC

Mine too since I added Theory to the mix with Atlas.
ID: 42002 · Report as offensive     Reply Quote
David Cameron
Project administrator
Project developer
Project scientist

Send message
Joined: 13 May 14
Posts: 366
Credit: 13,262,778
RAC: 6,995
Message 42013 - Posted: 31 Mar 2020, 7:47:57 UTC - in response to Message 42002.  

The max #CPUs per app is set in the project's plan_class_spec.xml file. It is 8 for vbox WU and 12 for native WU.

There are two servers sending WU - I can see that on one server this file was last updated on 23 March and on the other it is 17 March. So maybe there was an inconsistency between the two servers during those 6 days, but I'm not able to see the history of modifications to the file. I will ask the LHC admins if they know what happened.
ID: 42013 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 1557
Credit: 57,665,912
RAC: 203,128
Message 42299 - Posted: 27 Apr 2020, 8:00:12 UTC - in response to Message 42013.  
Last modified: 27 Apr 2020, 8:01:07 UTC

Credit for -native and Windows are the same.
One task is finished from AgileBoincers with -native and one with Windows.
https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=138245612
271904696 10452383 23 Apr 2020, 12:30:39 UTC 23 Apr 2020, 13:00:36 UTC Fehler beim Berechnen 18.02 0.00 --- ATLAS Simulation v2.00 (vbox64_mt_mcore_atlas) windows_x86_64
271909745 10548292 23 Apr 2020, 16:07:20 UTC 27 Apr 2020, 7:38:51 UTC Fertig und Bestätigt 8,840.36 29,760.67 373.09 ATLAS Simulation v2.00 (vbox64_mt_mcore_atlas) windows_x86_64
272013668 10635701 26 Apr 2020, 10:12:16 UTC 27 Apr 2020, 7:15:58 UTC Fertig und Bestätigt 4,071.31 43,248.55 373.09 ATLAS Simulation v2.82 (native_mt) x86_64-pc-linux-gnu
ID: 42299 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 1557
Credit: 57,665,912
RAC: 203,128
Message 43785 - Posted: 4 Dec 2020, 20:26:40 UTC

Since midday Credit-points are 30% from before.
-native and Windows.
ID: 43785 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 585
Credit: 33,205,523
RAC: 18,462
Message 43786 - Posted: 4 Dec 2020, 22:29:56 UTC

I see the same, but so far only for one Atlas task. Well maybe not 30%, but less than 50%. The device peak FLOPS was the same as for previous tasks which gave normal credit.
ID: 43786 · Report as offensive     Reply Quote
djoser
Avatar

Send message
Joined: 30 Aug 14
Posts: 145
Credit: 10,847,070
RAC: 0
Message 43789 - Posted: 5 Dec 2020, 8:37:07 UTC

I run ATLAS native and notice a drop in credits (>50%) every time the version number changes.
Why mine when you can research? - GRIDCOIN - Real cryptocurrency without wasting hashes! https://gridcoin.us
ID: 43789 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 1557
Credit: 57,665,912
RAC: 203,128
Message 43790 - Posted: 5 Dec 2020, 8:46:02 UTC - in response to Message 43789.  

Yes, this is when a new Application-Version is installed.
But this was in -native AND in Windows.
Now it seem back to normal for both.
ID: 43790 · Report as offensive     Reply Quote
Henry Nebrensky

Send message
Joined: 13 Jul 05
Posts: 158
Credit: 14,665,461
RAC: 0
Message 43807 - Posted: 9 Dec 2020, 15:44:10 UTC - in response to Message 43790.  

Yes, this is when a new Application-Version is installed.
But this was in -native AND in Windows.
Now it seem back to normal for both.
Having drafted this off-line: mine went down:

290583495 17,762.46 68,998.94 433.84 ATLAS Simulation v2.84 (native_mt) x86_64-pc-linux-gnu
290449999 17,692.59 69,222.89 473.77 ATLAS Simulation v2.83 (native_mt) x86_64-pc-linux-gnu
290381406 17,758.57 68,993.30 1,027.94 ATLAS Simulation v2.82 (native_mt) x86_64-pc-linux-gnu

There again, on the same machine I get
289610329 21,778.25 21,778.25 180.94 Theory Simulation v300.06 (native_theory) x86_64-pc-linux-gnu
290203308 32,569.85 32,259.48 298.90 SixTrack v502.05 (avx) x86_64-pc-linux-gnu
so the new rate of Atlas credits is much closer to that for Theory and SixTrack.

And having given it time to untangle itself, at last I got:
290642977 18,283.33 70,903.93 806.63 ATLAS Simulation v2.84 (native_mt) x86_64-pc-linux-gnu
290625445 17,840.59 69,256.80 315.02 ATLAS Simulation v2.84 (native_mt) x86_64-pc-linux-gnu
ID: 43807 · Report as offensive     Reply Quote

Message boards : ATLAS application : Change in Credit?


©2022 CERN