WTF happened here?
28 May 2007 12:14:04 UTC
Topic 192790
(moderation:
http://einsteinathome.org/workunit/33544459
0 credit for 16 hours of work!
Same thing one week before.
I am sure that the same thing is happening here:
http://einsteinathome.org/workunit/3349322
Is it a problem to crunch with AMD/Linux?
WTF happened here?
)
The first one listed shows:
Validate state Invalid
which means something happened to the computation to make it invalid (unuseable). If this is a pattern, then perhaps you might have an overheating issue, bad ram, excessive Overclock, etc. If it's an infrequent fluke, then I'd ignore it.
The second listed wu has already been purged/deleted, and is no longer available for examination.
RE: The first one listed
)
No overclocking, no bad RAM. Just a normal Athlon XP1700.
sorry link was wrong:
http://einsteinathome.org/workunit/33493228
RE: http://einstein.phys.uw
)
This seems to happen rather frequently when there are three copies of the same workunit sent to participants who have clients running under different operating systems. E.g. if two of them have OS A and one has OS B, then usually the validator will accept the two results from the "majority" OS A and invalidate that of OS B.
See the "Problems" section of the board, there are quite a few reports about this phenomenon.
CU
BRM
RE: RE: http://einstein.p
)
I just want my credits for this WU.
It can't be that this WU is invalid and claims no credits for this WU due to problems of the application.
Otherwise I have to spend my energy to other Projects like fight@aids. It is absolutly a shame that this bug is not fixed.
Ok jetzt in meiner
)
Ok jetzt in meiner Muttersprache!
Bernd es kann nicht sein, dass so eine Scheiße passeiert.
Die nexten WUs die anstehen brauchen 50 Std. und ich kann die eventuell wegschmeißem, weil idiotische Wintels anders rechnen?
Habt ihr überhaupt Tests auf die neue Application gemacht?
Langsam kotzt es mich wirklich an, wie ihr die energie verpuffen lasst.
RE: Ok jetzt in meiner
)
Let's get this into perspective:
It doesn't make any diffenerence , whatsoever, for the science part of this project. Whether 2 out of 3 or 3 out of 3 Results get validated is of no significance for the science as in both cases, the workunit is processed and a valid result is achieved. No energy is wasted. It's just the amount of "glass pebbles" paid that is different.
Auch mal in Deutsch: Ich kann nicht ganz nachvollziehen wie man sich ersthaft bei einem wissenschaftlichen Projekt dieses Kalibers darüber aufregen kann, dass man durch diesen Bug (oder was auch immer) pro Woche ein paar "Glasperlen" weniger gutgeschrieben bekommt. Das ist allenfalls ärgerlich, aber kein Grund sich aufzuregen, glaube ich. Hier geht's doch nur um virtuelle Credits, wenn Du willst kannst du gerne 150.000 von den Dingern von mir haben ;-)...
CU
BRM
RE: It doesn't make any
)
This worries me. I agree completely that the number of "glass pebbles" (or "brass washers", as we call them in the UK) awarded is completely meaningless. But what about the science?
A WU is issued to a Windows host and a Linux host. They disagree. It is issued, at random, to a third host to arbitrate. The OS of the third host determines which of the two previous results is 'valid', and hence which of the two results is entered into the science database as the "canonical" result. Surely that affects the scientific result of the experiment?
Are you saying that the variations between OSs are below the threshhold for scientific statistical reliability? I don't know that, and I'm fairly sure you don't know either. If they're insignificant, then it would be safe to relax the validation criteria: but that, in turn, would make a nonsense of the complaints about Akosf's optimisations at the start of the S5R1 run ("You have to use the approved, certified, application supplied by the project, because only the approved application will be accepted by the peer-review community as a guarantee of scientifically-valid results.").
Validation is supposed to validate the science. The awarding of credit is a side-effect of that. 'No credit' implies 'bad science' in this situation. And we base this on a lottery of host allocation? Something wrong here.
I agree with most of what
)
I agree with most of what you say, but I guess what has to be stressed is that validation is a statistical process that can fail in two ways:
False positives: Results are judged similar enough to validate, but really , on individual inspection, shouldn't (by any scientific rule you want)
False negatives: Results are judged not similar enough, but really none of them is really wrong.
In most situations like this you just can't minimize both, false positives and false negatives, at the same time. Clearly the interest of the scientists is to minimize false positives, and the credit-collectors would like to minimize false negatives.
I'm sure the guys in charge here are working on the problem, but I can understand that they don't go the easy way of relaxing the validator because that could hurt the science.
BRM
I've got a "Checked but no
)
I've got a "Checked but no consensus" WU that has already been sent to the third computer (Work unit ID 33818947). I've got a Mac OS X, the other computer is a Windows XP and the third computer is a Windows 2000. Given the previous comments on how the results from the majority OS are accepted, it will be interesting to see if the Windows machines are accepted and mine rejected.
I bet your computer sees
)
I bet your computer sees nothing whatsoever off the credits claimed.