rpi_logo
testing new credit policy on nbody workunits
testing new credit policy on nbody workunits
log in

Advanced search

Message boards : News : testing new credit policy on nbody workunits

Author Message
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0

Message 48344 - Posted: 29 Apr 2011, 19:55:28 UTC

I finally have BOINCs new credit policy implemented for the nbody workunits. It's to the point where I need to test it live. If you're having any problems with the nbody workunits being validated incorrectly, or being assigned weird amounts of credit (note -- the new credit policy is adaptive so the credit awarded can change over time, and it might take awhile to stabilize), this is the place to let me know.

Once I've gotten this debugged and working correctly, we'll have a days of double credit as promised for all our recent outages; and to make up for any weirdness in testing the new credit policy with the nbody workunits.

So I'd like to thank everyone for their patience in dealing with us getting everything updated and working correctly these last few weeks.

--Travis
____________

Zydor
Avatar
Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0

Message 48345 - Posted: 29 Apr 2011, 20:01:46 UTC
Last modified: 29 Apr 2011, 20:03:56 UTC

Can you give any yardstick/rough idea of credit levels set against a couple of machine types so we can guage if its wierd or not

Regards
Zy

Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0

Message 48346 - Posted: 29 Apr 2011, 20:06:14 UTC - in response to Message 48345.

Can you give any yardstick/rough idea of credit levels set against a couple of machine types so we can guage if its wierd or not

Regards
Zy


It should be similar to what it was awarding before -- what the clients were reporting for claimed credit (unfortunately this has been removed in the lastest versions of the server code so I don't have too much to check that against).
____________

Profile [AF>EDLS] Polynesia
Avatar
Send message
Joined: 5 Apr 09
Posts: 71
Credit: 6,120,786
RAC: 0

Message 48383 - Posted: 30 Apr 2011, 21:19:12 UTC

What is the amount of credit because, as the stats of the units is not on our behalf and we do not see anything ...
____________
Team Alliance francophone, boinc: 7.0.18

GA-P55-UD5, i7 860, Win 7 64 bits, 8g DDR3, GTX 470

Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0

Message 48425 - Posted: 2 May 2011, 3:22:10 UTC - in response to Message 48383.

I've also been testing the new credit policy over on DNA@Home, and it seems that it's on the rather low end of things as far as credit is concerned. I've added a credit modifier (which is currently set to 5x what the credit policy is returning).

If anyone can give me any info about how close the credit the nbody workunits are returning is to other projects, it would be much appreciated.
____________

Profile Beyond
Send message
Joined: 15 Jul 08
Posts: 383
Credit: 501,817,790
RAC: 0

Message 48430 - Posted: 2 May 2011, 14:34:44 UTC

Travis. Question, was this switch to the new credit scheme the "BIG NEWS"? Was afraid to ask at the time :(

Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0

Message 48439 - Posted: 2 May 2011, 19:54:09 UTC - in response to Message 48430.

Travis. Question, was this switch to the new credit scheme the "BIG NEWS"? Was afraid to ask at the time :(


I don't know if it's big news. But we're using BOINC's default credit system now, so things are definitely different.

Right now I'm just using it on the n-body application, as credit values for it are pretty difficult to calculate. I'm going to start testing it with the other application however -- but just to calculate the values initially (we'll still be using fixed credit), so I can compare how close BOINC's default estimated credit gets to what we're awarding with fixed credit. If we can get them close enough, I'll probably do the swap for the main application as well.
____________

Profile Beyond
Send message
Joined: 15 Jul 08
Posts: 383
Credit: 501,817,790
RAC: 0

Message 48446 - Posted: 2 May 2011, 23:06:41 UTC - in response to Message 48439.

I don't know if it's big news. But we're using BOINC's default credit system now, so things are definitely different.

Right now I'm just using it on the n-body application, as credit values for it are pretty difficult to calculate. I'm going to start testing it with the other application however -- but just to calculate the values initially (we'll still be using fixed credit), so I can compare how close BOINC's default estimated credit gets to what we're awarding with fixed credit. If we can get them close enough, I'll probably do the swap for the main application as well.

I was referring to the BIG NEWS COMING thread. think maybe it's disappeared now. Anyway, why change credit systems? Is there a reason? The 3 WU sizes are fixed, so fixed credit seems like the fair way to go. If you start throwing the totally brain dead BOINC flops estimates into the equation it'll not be pretty. If the new BOINC system is allowed to degrade the GPU credits downward compared to the same size CPU WUs it'll also be less than shiny. So far AFAIK the new system has been pretty much universally rejected/ignored. SETI might use it but they have their own reasons. Wondering?

Profile rkillian
Avatar
Send message
Joined: 4 Feb 10
Posts: 9
Credit: 51,176,789
RAC: 0

Message 48466 - Posted: 3 May 2011, 23:44:16 UTC - in response to Message 48439.



I don't know if it's big news. But we're using BOINC's default credit system now, so things are definitely different.

Right now I'm just using it on the n-body application, as credit values for it are pretty difficult to calculate. I'm going to start testing it with the other application however -- but just to calculate the values initially (we'll still be using fixed credit), so I can compare how close BOINC's default estimated credit gets to what we're awarding with fixed credit. If we can get them close enough, I'll probably do the swap for the main application as well.

Travis, If it is anything like the credit system they had a few months ago it will be bad news. The projects that I have seen that tried to use it the Credits dropped drasticly. Aqua credit dropped 90%. they had to abanden them and go back to their on system to prevent a mass exodus.

Royce
____________

Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0

Message 48467 - Posted: 4 May 2011, 5:15:43 UTC - in response to Message 48466.



I don't know if it's big news. But we're using BOINC's default credit system now, so things are definitely different.

Right now I'm just using it on the n-body application, as credit values for it are pretty difficult to calculate. I'm going to start testing it with the other application however -- but just to calculate the values initially (we'll still be using fixed credit), so I can compare how close BOINC's default estimated credit gets to what we're awarding with fixed credit. If we can get them close enough, I'll probably do the swap for the main application as well.

Travis, If it is anything like the credit system they had a few months ago it will be bad news. The projects that I have seen that tried to use it the Credits dropped drasticly. Aqua credit dropped 90%. they had to abanden them and go back to their on system to prevent a mass exodus.

Royce



Well if it sucks, we'll improve on it. :)
____________

Kenny Frew
Send message
Joined: 24 Feb 09
Posts: 5
Credit: 282,355
RAC: 0

Message 48469 - Posted: 4 May 2011, 11:20:19 UTC

Just thought I would mention, starting this morning(May 4) I have had 4 fails in a row. Windows Error - milkyway_nbody_0.40_windows_x86_64_mt.exe has stopped working. Computation Error.

Anton Rang
Send message
Joined: 25 Feb 11
Posts: 2
Credit: 1,353,635
RAC: 0

Message 48476 - Posted: 5 May 2011, 3:26:39 UTC - in response to Message 48469.

I'm seeing a bunch of failures for the nbody code the past day or two on MacOS due to the time limits being set far too low (3 seconds or so).

Profile Nekodemus vom Wolkenstein
Avatar
Send message
Joined: 28 Sep 10
Posts: 5
Credit: 46,734,494
RAC: 0

Message 48489 - Posted: 5 May 2011, 18:19:39 UTC

Hi Travis, I have the following problem.

MilkyWay @ home N-Body Simulation 0.40 (mt) 00:00:00 de_nbody_orphan_test_2model_4_68481_1304613093_2 planned duration.

The result:
aborting task de_nbody_orphan_test_2model_4_68481_1304613093_2: exeeded 0.000542 elapsed time limit.

Can you explain?
Thanks

Nekodemus
____________
regards
Nekodemus vom Wolkenstein

Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0

Message 48515 - Posted: 7 May 2011, 5:11:16 UTC - in response to Message 48489.

Hi Travis, I have the following problem.

MilkyWay @ home N-Body Simulation 0.40 (mt) 00:00:00 de_nbody_orphan_test_2model_4_68481_1304613093_2 planned duration.

The result:
aborting task de_nbody_orphan_test_2model_4_68481_1304613093_2: exeeded 0.000542 elapsed time limit.

Can you explain?
Thanks

Nekodemus


I've been on the BOINC mailing lists and having a bit of back and forth trying to track down this problem. I made a couple changes to the database today which hopefully should make it happen less often. Have you noticed any difference?
____________

scottishwebcamslive.com
Avatar
Send message
Joined: 10 Oct 07
Posts: 79
Credit: 69,337,972
RAC: 0

Message 48547 - Posted: 7 May 2011, 16:20:40 UTC

hi,

did i miss the double credit day ?

Ian
____________
....Please Join team Scotland HERE

Profile banditwolf
Avatar
Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0

Message 48548 - Posted: 7 May 2011, 16:31:55 UTC - in response to Message 48547.

hi,

did i miss the double credit day ?

Ian


No. I don't think it is near running well enough yet. Again today one 'fix' caused more problems. And still no cpu opti app.
____________
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

Profile Nekodemus vom Wolkenstein
Avatar
Send message
Joined: 28 Sep 10
Posts: 5
Credit: 46,734,494
RAC: 0

Message 48640 - Posted: 9 May 2011, 18:26:01 UTC - in response to Message 48515.

Hi Travis,

Give thanks for the reply. Published 2 days working the N-body simulation error free.

Thanks Nekodmus
____________
regards
Nekodemus vom Wolkenstein

John G
Send message
Joined: 1 Apr 10
Posts: 49
Credit: 171,863,025
RAC: 0

Message 48641 - Posted: 9 May 2011, 18:52:07 UTC

Server are down ???? FYI

Regards

John G

Profile Blurf
Volunteer moderator
Project administrator
Send message
Joined: 13 Mar 08
Posts: 804
Credit: 26,380,161
RAC: 0

Message 48662 - Posted: 9 May 2011, 23:46:29 UTC - in response to Message 48641.

Server are down ???? FYI

Regards

John G


The servers are up. Next time please post this in Number Crunching rather then the news forum. I'll see it sooner.

Thanks
____________


Post to thread

Message boards : News : testing new credit policy on nbody workunits


Main page · Your account · Message boards


Copyright © 2018 AstroInformatics Group