| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Stop TB
Thread: Ridiculous low credit claim for longer running tasks |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 99
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
oet1 127.76 pts/hour is a typo?? Or an artefact of short duration units?
Thanks for the work on this. I agree from my own limited viewpoint that it's the occasional significant outlier that grabs the attention. |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
It's interesting how uplinger's numbers compare to https://bit.ly/WCGCPH1 , which are quite a bit lower for the day and 7 day average.
----------------------------------------Edit: Frankly I'm at a loss, so revisited the data as extract using the xml pull and see uplinger: hst1 13.386 317.543 29.051971496746113 extract past 5 days: Date RunTime Points Results Credit/Hr Points divided by 7 which divided by (seconds divided by 3600) to get the hourly credit and still at a loss why this substantial difference. [Edit 2 times, last edit by SekeRob* at May 13, 2016 8:54:11 PM] |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
Hmm....My sql query was bonkers I guess...Thanks for the questions on my values.
----------------------------------------
Apparently: avg(r.granted_credit)/avg(r.cpu_time/3600.0) != avg(r.granted_credit/(r.cpu_time/3600.0)) I'm not sure if there is some major rounding issues or what happened, but this looks better, especially with OET1. But even with these being updated, the granted credit is still all sitting around 25 points per cpu, CEP2 is the high ball, but that has some interesting credit granting due to the jobs. Thanks, -Uplinger [Edit 1 times, last edit by uplinger at May 14, 2016 3:08:16 AM] |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Glad we landed on (my) same page, though actually had hoped something was [long standing] bonkers with mine and putting the hands on another 20% of the missing 80% extra that never arrived under Gen 3 Credit_New.
----------------------------------------The agreement over at Berkeley, CN is thoroughly broken... now the solution, which could be something of a 'per WU' mini benchmark used as base to compute the result credit [with some reality checks], WCG can set the value of the mini benchmark for each science, each science having a different calculation composition fpops/integer/io load  [Edit 1 times, last edit by SekeRob* at May 14, 2016 7:54:37 AM] |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Snip quote
During this investigation, I compared MCM1 credit per cpu hour for all results and graphed them. The bell curve between the granted credit on HST and MCM looked familiar. This to me drew me to the conclusion that it is mostly the outlier credit granted that needed to be investigated. From this, I looked into why a few workunits, less than 20 per day were coming back with a low claimed credit. This is where I spent most of my time. I have tweaked the credit granted algorithm to correct for this issue and to have a more normalized credit granted to members for the HST1 project. This has been installed with the updated HST1 validator that was installed a few hours ago. Thanks, -Uplinger The credit/hour trend has been on the up-up a few weeks and certainly appears to rise on in the past 5 days, to now show a day credit average of 25.92 and a rolling week mean of 25.06, closing in on MSM and UGM [which has very stable runtimes, breaking with the variability off from the runtime averages undulation. See https://bit.ly/WCGHST1 [red mini chart right bottom highlights credit changes] and https://bit.ly/WCGCPH1 [:thumbsup smiley] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Credit trends at WCG are like leaves on a tree... depending on wind direction you see the top or the bottom. Here.'s a riddle, where riddicule are not far apart in the English dictionar
----------------------------------------Result Name OS type OS version App Version Number Status Sent Time Time Due / Return Time CPU Time / Elapsed Time (hours) Claimed/ Granted BOINC Credit HST1_ 003078_ 000049_ AC0003_ T350_ F00049_ S00005_ 1-- Microsoft Windows 8.1 Professional x64 Edition, (06.03.9600.00) 714 Valid 5/31/16 10:33:11 6/2/16 18:03:07 21.28 462.1 / 300.5 HST1_ 003078_ 000049_ AC0003_ T350_ F00049_ S00005_ 0-- Microsoft Windows 8.1 Core x64 Edition, (06.03.9600.00) 714 Valid 5/31/16 10:33:09 6/8/16 10:07:52 16.59 241.2 / 300.5 My device normally gets a claim assignment of about 350 for these T350, but today they were good for the 240-250 range. Then the Wingman gets 462 assigned, albeit that was 8 days ago, and neither get a grant of half of sum, but something of a whim. Anyone who can make sense of this? (Dont try, it causes braincells to explode, a TIA maybe, Temporary Interruption of Airflow) [Edit 1 times, last edit by Former Member at Jun 8, 2016 12:48:15 PM] |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: 
|
Credits are completely insane at both ends of the spectrum. WCG represents the low end and on the other end there is Bitcoin Utopia. For ~$55 USD you can buy a USB ASIC miner and get 33,000 credits every 2.5 minutes. For less than the price of a GPU you can build an array in a powered hub and get a half billion credits/week.
----------------------------------------   Credits are meaningless but are the only way one can compare participation. There needs to be some regulation but it will never happen as long as the BOINC app is open source and can be manipulated by any project. Cheers, Rob ... always good to hear from you.   [Edit 3 times, last edit by Dataman at Jun 8, 2016 4:29:28 PM] |
||
|
|
BobCat13
Senior Cruncher Joined: Oct 29, 2005 Post Count: 295 Status: Offline Project Badges: 
|
Anyone who can make sense of this? Not really. The granted credit is lower than both of the claimed credits. Project Name: Help Stop TB Created: 06/08/2016 17:03:04 Name: HST1_003518_000018_MC0006_T300_F00066_S00005 Minimum Quorum: 2 Replication: 2 Result Name OS type OS version App Version Number Status Sent Time Time Due / Return Time CPU Time / Elapsed Time (hours) Claimed/ Granted BOINC Credit HST1_ 003518_ 000018_ MC0006_ T300_ F00066_ S00005_ 1-- Linux 3.16.0-38-generic 714 Valid 6/8/16 17:03:10 6/10/16 05:16:36 13.75 401.7 / 399.5 HST1_ 003518_ 000018_ MC0006_ T300_ F00066_ S00005_ 0-- Linux 3.16.0-4-amd64 714 Valid 6/8/16 17:03:10 6/9/16 10:34:25 10.13 475.0 / 399.5 |
||
|
|
scol5913
Senior Cruncher UK Joined: May 30, 2013 Post Count: 192 Status: Offline Project Badges:
|
Over the last few days my pts/hr have swung from 130 up to 180 and back to 155. Can the WUs be that different, its all very odd. Maybe tomorrow 200.
|
||
|
|
supdood
Senior Cruncher USA Joined: Aug 6, 2015 Post Count: 333 Status: Offline Project Badges: 
|
My favorite of my WUs so far... 188.4 credit claim for 24 hours vs 127.5 for 3.8 on an i7 5600U 3 days apart?
----------------------------------------HST1_ 003793_ 000084_ AT0020_ T001_ F00043_ S00004_ 0-- Microsoft Windows 7 Enterprise x64 Edition, Service Pack 1, (06.01.7601.00) 714 Pending Validation 6/13/16 16:04:10 6/13/16 20:11:24 3.80 127.5 / 0.0 HST1_ 003784_ 000019_ KC0016_ T325_ F00043_ S00005_ 0-- Microsoft Windows 7 Enterprise x64 Edition, Service Pack 1, (06.01.7601.00) 714 Pending Validation 6/13/16 12:33:18 6/16/16 14:56:37 24.04 188.4 / 0.0 EDIT: Had another WU in this group come through with 22.98 188.4 / 576.2 (higher than wingman claimed as well) for an average credit/hour for this machine. ---------------------------------------- [Edit 2 times, last edit by supdood at Jun 16, 2016 3:44:44 PM] |
||
|
|
|