Experiment for server operations check...

Message boards : News : Experiment for server operations check...
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 5 · Next

AuthorMessage
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 26,981,856
RAC: 717
United States
Message 43643 - Posted: 21 Aug 2012, 2:08:10 UTC


We're going to be doing an experiment to see if we understand what caused the bad work estimates and to bad credits. Unfortunately the test will replicate the conditions that caused the problems.

You know, of course that I'll get around to fixing your credit when we're done.
ID: 43643 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43658 - Posted: 25 Aug 2012, 9:35:03 UTC - in response to Message 43643.  


We're going to be doing an experiment to see if we understand what caused the bad work estimates and to bad credits. Unfortunately the test will replicate the conditions that caused the problems.

You know, of course that I'll get around to fixing your credit when we're done.


Oh hard times :-) The experiment is certainly running. 6.04's for all my computers went from having gone down to reasonable estimates (almost perfect) to levels where "oh my god" would be appropriate to say...

Acer with HD4850 now est one 6.04 to 17 hours, 38 minutes /WU (4-6 hours is normal depending on blanking), Samsung Laptop with Nvidia GeForce 315M now est one 6.04 to 30 days 12 hours (8-9 hours is normal), and my poor little ION now est one 6.04 to 3 days 8 hours (24-28 hours is normal)

But there's no panic here, let the experiment continue.
ID: 43658 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43659 - Posted: 25 Aug 2012, 11:46:54 UTC
Last modified: 25 Aug 2012, 12:27:21 UTC

I forgot to add, when I still could edit the post:

As far as credit is concerned, it is as expected a bit flakey to say the least. Everything in between 20K to just only 4 (not K) is given in credit for a 6.04.

Still nothing that bothers me the least. After all this is Beta, and I do not expect or demand any credit at all really. Everything above 0 is a great bonus.

Edit, added: Just noticed that a bit of experimentation seems to be going on with the 0.xx CPU + 1.00 ATI or Nvidia settings too.
ID: 43659 · Report as offensive
B-Man
Volunteer tester

Send message
Joined: 24 Aug 09
Posts: 79
Credit: 26,117
RAC: 0
United States
Message 43660 - Posted: 25 Aug 2012, 20:05:52 UTC

Ok you are getting CPU back for the server test. I had set Beta to GPU only while we wait for a GPU science app for OSX. May be a few hours but I will grab some more.
ID: 43660 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43661 - Posted: 26 Aug 2012, 15:03:14 UTC

We're certainly not getting anywhere fast with "Number of tasks completed", when 90% or even more of all tasks are so heavily blanked like they are now. Getting one <10% blanked seems nearly impossible. Most are now >60% blanked, and >80% is not uncommon.

Maybe that is part of the experiment, what do I know?

ID: 43661 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 26,981,856
RAC: 717
United States
Message 43662 - Posted: 26 Aug 2012, 21:19:14 UTC - in response to Message 43661.  

It's not part of the experiment and it's really slowing things down. It's hard to know in advance what the faction blanked is going to be for a data tape.
ID: 43662 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43664 - Posted: 27 Aug 2012, 19:11:34 UTC - in response to Message 43662.  

It's not part of the experiment and it's really slowing things down. It's hard to know in advance what the faction blanked is going to be for a data tape.


To me at least, the problem is minimal. I'm only 56 yo, and judging my how old my father is (95), and how old most of my relatives have been when they died, I expect to be here for a considerable time :-)

I can wait....
ID: 43664 · Report as offensive
Profile cAnDYmanS@H-Beta
Volunteer tester
Avatar

Send message
Joined: 24 May 12
Posts: 38
Credit: 436,379
RAC: 0
Romania
Message 43676 - Posted: 31 Aug 2012, 6:28:05 UTC

I am also seeing this great swing from WUs being validated for more than 20k for no apparent reason, while others get less than 1 credit. However I did also have a unit that got 692 which is kind of funny since errored out WUs get 700 and they only need to run for a couple of minutes instead of the full cycle.

But I've learned my lesson, so as Sten-Arne said: not complaining, just trying to help by pointing out potential issues.

Cheers!
Per aspera, ad astra!

ID: 43676 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43679 - Posted: 1 Sep 2012, 10:47:07 UTC

A credit of 0.79-0.89 seems to be the new normal for a 6.04, on my Q8200 with the ATI HD4850 card. Perhaps just a tiny, tiny bit on the low side of what I would call normal. :-)

It's OK though, I still do not complain. I'm just observing and smiling.
ID: 43679 · Report as offensive
S@NL - John van Gorsel
Volunteer tester

Send message
Joined: 26 Jun 10
Posts: 22
Credit: 586,393
RAC: 56
Netherlands
Message 43680 - Posted: 1 Sep 2012, 14:45:18 UTC

I just received 0.2 credits for 74,078.92 seconds of work on an AP task on my Q9650. Two days ago I received not less then 20 credits for the same amount of work.

ID: 43680 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43681 - Posted: 1 Sep 2012, 16:31:28 UTC - in response to Message 43680.  

I just received 0.2 credits for 74,078.92 seconds of work on an AP task on my Q9650. Two days ago I received not less then 20 credits for the same amount of work.


Bad economic times, with hyper inflation, may be the simple answer to the loss of credit values :-)

It will come better times, it always does...

LOL
ID: 43681 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 3
United Kingdom
Message 43682 - Posted: 1 Sep 2012, 17:17:36 UTC - in response to Message 43680.  
Last modified: 1 Sep 2012, 17:19:41 UTC

I just received 0.2 credits for 74,078.92 seconds of work on an AP task on my Q9650. Two days ago I received not less then 20 credits for the same amount of work.

You failed to mention that your wingman's GPU only took 250.23 seconds to do that Wu, and that most his GPU Wu's take less than 1,000 seconds, and a lot of them are inconclusive or invalid, You were just unlucky that your result validated with his.

Computer 57035

Claggy
ID: 43682 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 26,981,856
RAC: 717
United States
Message 43684 - Posted: 2 Sep 2012, 13:38:35 UTC - in response to Message 43682.  

Low credit is part of the problem we're trying to debug. Expect it to continue. Credit will be fixed once were sure we understand the problem and have it fixed.
ID: 43684 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43685 - Posted: 2 Sep 2012, 15:51:46 UTC - in response to Message 43684.  

Low credit is part of the problem we're trying to debug. Expect it to continue. Credit will be fixed once were sure we understand the problem and have it fixed.


Well Dr Korpela, at least I can confirm that the experiment does not involve any kind of physical pain :-)
ID: 43685 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43689 - Posted: 5 Sep 2012, 15:16:33 UTC
Last modified: 5 Sep 2012, 15:17:48 UTC

Still no pain whatsoever, only low credits on all my computers. It's getting more and more consistenly very low, the more WU's I finish. In a short while if it continues like this, I will get negative credits :-)

Are we getting anywhere with the solution? I'm in no hurry, just curious.
ID: 43689 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 26,981,856
RAC: 717
United States
Message 43690 - Posted: 6 Sep 2012, 3:34:29 UTC - in response to Message 43689.  

Still working on the "why" of the problem and we don't yet have enough statistic to know whether it's a small number of low credit claims that are getting amplified somehow, or if there's something else going on.

David and I will be going through a lot of scheduler and validator logs next week...
ID: 43690 · Report as offensive
Profile cAnDYmanS@H-Beta
Volunteer tester
Avatar

Send message
Joined: 24 May 12
Posts: 38
Credit: 436,379
RAC: 0
Romania
Message 43691 - Posted: 6 Sep 2012, 6:29:52 UTC

I can confirm that all of the WUs that I have completed in the past few days get undercredited with less than 2 credits or even 0.00. On the bright side, I don't see WUs getting too much credit anymore (over 5k).

Hope this helps. Cheers!
Per aspera, ad astra!

ID: 43691 · Report as offensive
Father Ambrose
Volunteer tester

Send message
Joined: 1 May 07
Posts: 556
Credit: 6,441,065
RAC: 370
United Kingdom
Message 43692 - Posted: 6 Sep 2012, 10:00:41 UTC

Just been looking through valid WU,s most are still high credits.

Low credits appear to be paired with CUDA WU,s the valid credits being same as
wingman LOWEST and not wingman HIGHEST reported credits

Might be worth looking at.

Michael
ID: 43692 · Report as offensive
Celeron M processor 1.50GHz
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1659
Credit: 12,789,696
RAC: 7,691
Sweden
Message 43693 - Posted: 6 Sep 2012, 15:28:37 UTC - in response to Message 43690.  

Still working on the "why" of the problem and we don't yet have enough statistic to know whether it's a small number of low credit claims that are getting amplified somehow, or if there's something else going on.

David and I will be going through a lot of scheduler and validator logs next week...


Thanks for the update Eric.

Take your time, I'm not here for the credit, neither am I on main for the credit, which can be seen by me having moved 99% of my resources over to Beta, in order to let you get as many results back as possible, for your analyzis of the problems.
ID: 43693 · Report as offensive
Profile cAnDYmanS@H-Beta
Volunteer tester
Avatar

Send message
Joined: 24 May 12
Posts: 38
Credit: 436,379
RAC: 0
Romania
Message 43694 - Posted: 6 Sep 2012, 17:29:14 UTC - in response to Message 43693.  

I'm right behind you, Sten-Arne! I too will keep the fire burning under this old pot.

Keep 'em coming, Eric!
Per aspera, ad astra!

ID: 43694 · Report as offensive
1 · 2 · 3 · 4 . . . 5 · Next

Message boards : News : Experiment for server operations check...


 
©2019 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.