Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Completed Research Forum: The Clean Energy Project - Phase 2 Forum Thread: credit for WU's that error out |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 9
|
Author |
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I currently have 4 WU's that have errored out - for a loss of computer time of 54.76 hours - Is it possible to get some credit for all these errored WU's?
(1) E225118_ 582_ S.338.C44H32N6.MCLMFQSZGQLIMW-UHFFFAOYSA-N.3_ s1_ 14_ 0-- M-09 Error 9/2/14 17:30:03 9/6/14 12:31:34 8.72 / 8.75 200.2 / 0.0 (2) E225108_ 948_ S.328.C43H31N1S2.QXKUWPGSVZUAMU-UHFFFAOYSA-N.8_ s1_ 14_ 0-- M-09 Error 8/29/14 18:42:46 9/4/14 17:07:33 10.04 / 10.05 310.3 / 0.0 M-09 - 2 Error, 2 Valid, 1 Pending Verification, 3 In Progress -- total 8 Wu's (25% error rate) (3) E225125_ 649_ S.348.C45H29N7.DXYVKCVCYJJGQU-UHFFFAOYSA-N.1_ s1_ 14_ 0-- M-08 Error 9/2/14 22:56:14 9/8/14 15:02:46 18.00 / 18.21 124.2 / 0.0 (4) E225119_ 148_ S.340.C44H32N4S1.CDXUISHLXPVDSI-UHFFFAOYSA-N.11_ s1_ 14_ 1-- M-08 Error 9/2/14 17:40:59 9/7/14 16:10:23 18.00 / 18.20 130.1 / 0.0 M-08 - 2 Error, 2 In Progress -- total 4 Wu's (50% error rate) 4 failed WU's - 54.76 total time lost |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The ethos is WCG was always that you'd get no credit for a WU in error if that error was caused by your machine, but you'd get at least some sort of credit (perhaps only half what you claimed) if the error was not your fault; for example if the WU was badly constructed or the error was caused by some sort of out-of-range condition in the modelling software.
It is not clear to me that that is still happening in the current CEP2 environment. I made a similar comment during the recent beta test and I'm disappointed that the techs have not commented. Unfortunately we don't necessarily get to see all the information they do, so we mustn't judge too harshly as to the usefulness of what we send back, but I still think the techs should keep us better informed about the current issues and do all they can to credit us for no-fault errors. |
||
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1664 Status: Offline Project Badges: |
CEP2: wasted time, energy, hardware resources, and finally ... wasted volunteer's good will !
----------------------------------------Yves |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I seem to be getting some Errors credited...
E225145... NX Error 05/09/14 15:02:00 08/09/14 03:00:46 18.00 / 18.09 560.8 / 560.8 E225119... Constitution Error 03/09/14 18:05:50 06/09/14 15:02:56 9.67 / 10.00 348.6 / 348.6 I currently have 8 errors listed, and 11 pages of valids, so obviously still a few problems but largely OK. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The script to 'credit' the 'not the volunteers machine's fault' to my contributions does not run continuously. Same as with the 'too late'. These take outs are set aside and reviewed as and when.
Cleanenergy mentioned an imminently monthly conference call where he'd put another specific discussion point up, the server status page. No word back at all is no good, tests the patience, but is it truly our worry? Yes and no. It's so easy to mod the validator rules and we'd and much more importantly, the scientists be nothing the wiser, so these errors being spat out is better i'd think till they can put the finger on the weak point. Individually we can make out choice to help ferret these issues out, or move to a science which does only a fail 1:1000, and then still a post is made to say 'my machine is rock solid, how come'. Me too , but do research the forums first if the issue was already being reported and how often. My 2 vatus |
||
|
Werinbert
Advanced Cruncher United States Joined: Jun 13, 2013 Post Count: 56 Status: Offline Project Badges: |
Could some one then explain this set of granting of credits:
----------------------------------------E225149_ 903_ S.382.C52H42N4.QXOQOLWJKMTFLQ-UHFFFAOYSA-N.14_ s1_ 14_ 3-- 640 Valid 9/13/14 07:24:11 9/13/14 15:06:34 3.33 68.5 / 68.5 Mine was the last one in the list. I returned something valid yet I got very few points for the work done. Another cruncher turned in an invalid result and got much more credits. following this logic it would be better for me to produce errors then actually produce valid results. |
||
|
ca05065
Senior Cruncher Joined: Dec 4, 2007 Post Count: 325 Status: Offline Project Badges: |
20:19 04th Sept - _0 sent out
----------------------------------------20:20 04th Sept - _1 sent out 20:28 04th Sept - _1 returns as error 20:33 04th Sept - _2 sent out 12:06 05th Sept - _0 returned 07:23 13th Sept - _2 returned Assume disagreement and both _0 and _2 put into pending verificaion. 07:24 13th Sept - _3 sent out 15:06 13th Sept - _3 returned Results then assigned as valid / invalid The allocation of points follows some rules which are not always clear. I am not an expert but the following is my understanding: The simplest and most usual method is to average the two claims. If the validator considers one claim to be 'out of normal' then it uses the other claim. CEP2 has an extra rule which only allows claims for work completed. My assumption is that the valid work units both failed in job 0 so do not claim much while the invalid work unit ran for a long time (over 15 hours) through to job 6 or job 7 and made a high claim. As _2 is invalid it received half its claim. _0 is considered 'out of normal' so _3 claim is used. [Edit 1 times, last edit by ca05065 at Sep 14, 2014 7:29:20 PM] |
||
|
Werinbert
Advanced Cruncher United States Joined: Jun 13, 2013 Post Count: 56 Status: Offline Project Badges: |
ca05065 thanks for the post.
----------------------------------------Unfortunately this makes the WCG granting of credits seem arbitrary. Especially when if I had gotten the invalid result my credits would have been about 3x greater than what I had actually been granted for returning a valid result. Due to oddities like this I am no longer using my high end machines to crunch for WCG rather I am letting WCG crunch on my slow machines where the time is more relevant than credits. Edit: I believe the task continued until Job#6. Also I took a look at the last task remaining on my computer (different from previous). It errored and I got a mere 138/138 (asking/granted) credits, the other wingmen also errored but they got around 450/450 credits. Again making WCG look very arbitrary. [Edit 1 times, last edit by Werinbert at Sep 15, 2014 8:59:07 AM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There's always some that will jump in that credits/points are subordinate, yet it is the first if not second most frequently discussed issue on these forums as source of irritation. Credit/points actually is the base of calculating of how much computing power you are contributing, where time is just time, even if you attach an atari running 24/24.
The asking/claim is pretty fake. It's what the servers compute based on a number of statistical averages for the project and your device performance benchmark. Say latter weighs in 20 percent of the total 'claim'. Read somewhere it was supposed to be removed altogether. Then you get to see 68.5 only, for using 10 times as much resources as a faah job does. Never been able to wrap my head around that. |
||
|
|