What the heck has happen to workunit credits?


log in

Advanced search

Message boards : Number crunching : What the heck has happen to workunit credits?

1 · 2 · Next
Author Message
Keith White
Avatar
Send message
Joined: 29 May 99
Posts: 370
Credit: 2,775,830
RAC: 2,142
United States
Message 1239427 - Posted: 1 Jun 2012, 13:28:25 UTC

Suddenly I'm getting squat credits (4-13) for workunits that use to give me 80 or more (based on how long it takes to crunch). I don't check in here as often as I should but did something recently (last 3 months) change in how it gets calculated?

In both of these cases the workunit was crunched by the same wingman using "SETI@home Enhanced Anonymous platform (NVIDIA GPU)" on their GTX 590.

So is it something wonky with his cruncher or did the credit calculations get tweaked again?
____________
"Life is just nature's way of keeping meat fresh." - The Doctor

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 4086
Credit: 111,822,495
RAC: 147,452
United States
Message 1239429 - Posted: 1 Jun 2012, 13:31:59 UTC

A year, or was it two, they changed things with CreditNew. Which has its ups and down. There are a few topics on the subject.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8465
Credit: 48,987,573
RAC: 74,019
United Kingdom
Message 1239431 - Posted: 1 Jun 2012, 13:41:17 UTC - in response to Message 1239427.

In both of these cases the workunit was crunched by the same wingman using "SETI@home Enhanced Anonymous platform (NVIDIA GPU)" on their GTX 590.

Morten Ross is running an extremely recent 'Multibeam x41x Preview, Cuda 4.20' test application on all six GTX 590 devices in his Longhorn server. I think this must be a hand-picked test unit - hopefully it is being closely monitored (not least, for side-effects of the current server bug).

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 23820
Credit: 32,641,631
RAC: 23,458
Germany
Message 1239434 - Posted: 1 Jun 2012, 13:51:43 UTC - in response to Message 1239431.

In both of these cases the workunit was crunched by the same wingman using "SETI@home Enhanced Anonymous platform (NVIDIA GPU)" on their GTX 590.

Morten Ross is running an extremely recent 'Multibeam x41x Preview, Cuda 4.20' test application on all six GTX 590 devices in his Longhorn server. I think this must be a hand-picked test unit - hopefully it is being closely monitored (not least, for side-effects of the current server bug).


Agreed.
Just checked another cuda alpha host with same result.
10-30% causes in low credit.

____________

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8465
Credit: 48,987,573
RAC: 74,019
United Kingdom
Message 1239443 - Posted: 1 Jun 2012, 14:10:05 UTC - in response to Message 1239434.

In both of these cases the workunit was crunched by the same wingman using "SETI@home Enhanced Anonymous platform (NVIDIA GPU)" on their GTX 590.

Morten Ross is running an extremely recent 'Multibeam x41x Preview, Cuda 4.20' test application on all six GTX 590 devices in his Longhorn server. I think this must be a hand-picked test unit - hopefully it is being closely monitored (not least, for side-effects of the current server bug).

Agreed.
Just checked another cuda alpha host with same result.
10-30% causes in low credit.

Well, I'm running the same application myself (which is how I know how new it is), and I don't see any sign of lowered credits on my hosts. I don't think it's an application issue, but I wondered if the overall performance of BOINC on a host with that many devices was being monitored.

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 23820
Credit: 32,641,631
RAC: 23,458
Germany
Message 1239500 - Posted: 1 Jun 2012, 16:21:10 UTC - in response to Message 1239443.

In both of these cases the workunit was crunched by the same wingman using "SETI@home Enhanced Anonymous platform (NVIDIA GPU)" on their GTX 590.

Morten Ross is running an extremely recent 'Multibeam x41x Preview, Cuda 4.20' test application on all six GTX 590 devices in his Longhorn server. I think this must be a hand-picked test unit - hopefully it is being closely monitored (not least, for side-effects of the current server bug).

Agreed.
Just checked another cuda alpha host with same result.
10-30% causes in low credit.

Well, I'm running the same application myself (which is how I know how new it is), and I don't see any sign of lowered credits on my hosts. I don't think it's an application issue, but I wondered if the overall performance of BOINC on a host with that many devices was being monitored.


You have some as well.

http://setiathome.berkeley.edu/result.php?resultid=2461040880
____________

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8465
Credit: 48,987,573
RAC: 74,019
United Kingdom
Message 1239503 - Posted: 1 Jun 2012, 16:30:34 UTC - in response to Message 1239500.

In both of these cases the workunit was crunched by the same wingman using "SETI@home Enhanced Anonymous platform (NVIDIA GPU)" on their GTX 590.

Morten Ross is running an extremely recent 'Multibeam x41x Preview, Cuda 4.20' test application on all six GTX 590 devices in his Longhorn server. I think this must be a hand-picked test unit - hopefully it is being closely monitored (not least, for side-effects of the current server bug).

Agreed.
Just checked another cuda alpha host with same result.
10-30% causes in low credit.

Well, I'm running the same application myself (which is how I know how new it is), and I don't see any sign of lowered credits on my hosts. I don't think it's an application issue, but I wondered if the overall performance of BOINC on a host with that many devices was being monitored.

You have some as well.

http://setiathome.berkeley.edu/result.php?resultid=2461040880

Interestingly, my wingmate (the valid one) on that one was another server OS. That's one of the major weaknesses of 'CreditNew' - we can no longer see which member of the quorum pair is making the low claim, which makes it much harder to analyse whether a host has a consistent problem.

But thanks - I'll keep an eye on it.

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 23820
Credit: 32,641,631
RAC: 23,458
Germany
Message 1239507 - Posted: 1 Jun 2012, 16:52:53 UTC

Look at this example no Windows server involved.

http://setiathome.berkeley.edu/workunit.php?wuid=997649948
____________

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 23820
Credit: 32,641,631
RAC: 23,458
Germany
Message 1239515 - Posted: 1 Jun 2012, 17:31:30 UTC

I think the servers messed up.

Look at reported times and credits.

http://setiathome.berkeley.edu/workunit.php?wuid=998136901
____________

Profile Area 51
Avatar
Send message
Joined: 31 Jan 04
Posts: 965
Credit: 42,193,520
RAC: 0
United Kingdom
Message 1239519 - Posted: 1 Jun 2012, 17:37:46 UTC - in response to Message 1239515.

I think the servers messed up.

Look at reported times and credits.

http://setiathome.berkeley.edu/workunit.php?wuid=998136901



Good spot. At least is shows that x41x isn't the only application demonstrating these characteristics. Question is, how widespread is this issue......
____________

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 23820
Credit: 32,641,631
RAC: 23,458
Germany
Message 1239525 - Posted: 1 Jun 2012, 17:45:29 UTC - in response to Message 1239519.

I think the servers messed up.

Look at reported times and credits.

http://setiathome.berkeley.edu/workunit.php?wuid=998136901



Good spot. At least is shows that x41x isn't the only application demonstrating these characteristics. Question is, how widespread is this issue......


At least all wingmen were cuda.

Not one ATI result messed up.

____________

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8465
Credit: 48,987,573
RAC: 74,019
United Kingdom
Message 1239568 - Posted: 1 Jun 2012, 18:21:30 UTC - in response to Message 1239525.

I think the servers messed up.

Look at reported times and credits.

http://setiathome.berkeley.edu/workunit.php?wuid=998136901

Good spot. At least is shows that x41x isn't the only application demonstrating these characteristics. Question is, how widespread is this issue......

At least all wingmen were cuda.

Not one ATI result messed up.

That particular case is exactly the one I wrote up on Monday in Host falling back to CPU processing... (the thread title is from an earlier mis-diagnosis).

It comes - or came then - when you report more than 64 tasks at once, and request new work at the same time.

Maybe no ATI cards could complete 64 tasks during the outage? :P

Profile ignorance is no excuse
Avatar
Send message
Joined: 4 Oct 00
Posts: 9529
Credit: 44,433,274
RAC: 0
Korea, North
Message 1239595 - Posted: 1 Jun 2012, 19:12:42 UTC - in response to Message 1239568.

I'd like ot know why I got over 2000WU's after the previous outage. Suddenly using BOINC 7.0.28 gives me an enormous cache.
____________
In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

End terrorism by building a school

Profile [B^S] madmac
Volunteer tester
Avatar
Send message
Joined: 9 Feb 04
Posts: 1139
Credit: 3,657,579
RAC: 4,019
United Kingdom
Message 1239600 - Posted: 1 Jun 2012, 19:20:05 UTC

I used to get around 90 for each WU now with my WU with either Fermi or Cuda I only get 30 so I have lost two thirds of my credits I have only noticed this since the outage
____________

Sten-Arne
Volunteer tester
Send message
Joined: 1 Nov 08
Posts: 3406
Credit: 19,632,252
RAC: 18,336
Sweden
Message 1239612 - Posted: 1 Jun 2012, 19:58:41 UTC - in response to Message 1239515.
Last modified: 1 Jun 2012, 20:03:49 UTC

I think the servers messed up.

Look at reported times and credits.

http://setiathome.berkeley.edu/workunit.php?wuid=998136901


Now that's a weird one. Run time was 323.00 Seconds, but CPU time was 6,747.00 seconds, still it was finished and uploaded in just over 5 minutes.

Sent from SETI 1 Jun 2012 | 1:10:18 UTC

Time reported: 1 Jun 2012 | 1:15:41 UTC


It's definitely something very strange going on there.

Edit: Just noticed Richard's explanation for this one. Consider this entire message of mine, as a waste of bandwidth.
LOL
____________

Profile Area 51
Avatar
Send message
Joined: 31 Jan 04
Posts: 965
Credit: 42,193,520
RAC: 0
United Kingdom
Message 1239616 - Posted: 1 Jun 2012, 20:04:29 UTC - in response to Message 1239600.

I used to get around 90 for each WU now with my WU with either Fermi or Cuda I only get 30 so I have lost two thirds of my credits I have only noticed this since the outage


I think we are definitely in the realm of unintended consequences.....
____________

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 23820
Credit: 32,641,631
RAC: 23,458
Germany
Message 1239680 - Posted: 1 Jun 2012, 21:33:23 UTC - in response to Message 1239568.

I think the servers messed up.

Look at reported times and credits.

http://setiathome.berkeley.edu/workunit.php?wuid=998136901

Good spot. At least is shows that x41x isn't the only application demonstrating these characteristics. Question is, how widespread is this issue......

At least all wingmen were cuda.

Not one ATI result messed up.

That particular case is exactly the one I wrote up on Monday in Host falling back to CPU processing... (the thread title is from an earlier mis-diagnosis).

It comes - or came then - when you report more than 64 tasks at once, and request new work at the same time.

Maybe no ATI cards could complete 64 tasks during the outage? :P


I reported 587 at once yesterday without a problem.
scratching head.

____________

Grant (SSSF)
Send message
Joined: 19 Aug 99
Posts: 5792
Credit: 58,070,261
RAC: 48,385
Australia
Message 1239694 - Posted: 1 Jun 2012, 21:54:45 UTC - in response to Message 1239680.

I reported 587 at once yesterday without a problem.
scratching head.

I was reporting hundreds of WUs at a time yesterday, but only 64 would actually be reported with each contact with the Scheduler.
And luckily at the time there were hundreds of WUs still uploading while i was hammering the update button so i didn't request new work while trying to report.

____________
Grant
Darwin NT.

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8465
Credit: 48,987,573
RAC: 74,019
United Kingdom
Message 1239699 - Posted: 1 Jun 2012, 22:14:21 UTC - in response to Message 1239680.

I reported 587 at once yesterday without a problem.
scratching head.

Check - count, even - the number of

28-May-2012 10:11:47 [SETI@home] [sched_op] handle_scheduler_reply(): got ack for task 16dc10ac.2114.14009.12.10.195_0

you get in reply to your 587. You run <sched_op_debug>, I'm sure? It's the only way to keep an eye on what's happening round here.

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 23820
Credit: 32,641,631
RAC: 23,458
Germany
Message 1239701 - Posted: 1 Jun 2012, 22:23:26 UTC

Sure i do.

I report more than 70 twice a day (at least).
Only on GPU of course.

____________

1 · 2 · Next

Message boards : Number crunching : What the heck has happen to workunit credits?

Copyright © 2014 University of California