Computer permanently claims 0 credits
23 Dec 2005 13:55:55 UTC
Topic 190431
(moderation:
What's up with this machine?
http://einsteinathome.org/host/335477
The last 200 results were delivered with state "success" and 0 credits claimed (and 0 granted). Most/A lot of these results were sent back to the server only a few minutes after they were received - there can't have been much work done.
As the results say "success", they are used for the calculation of the granted credits and bring them down a little bit, I think, because the lower of the next two results is used for the granted credit. The machine itself gets 0 credits granted.
Is anyone (forum administrators?) able to contact the anonymous owner? Or what else should be done?
Happy holidays,
GT
Computer permanently claims 0 credits
)
If you look, that 0 is not affecting the scores. Yes it is taking longer for some of the WUs to get done, cause this person is sending back garbage, but it's not hurting scores or science. It takes 3 valid results to make a quorum, and all the 0 ones he has has 3 valid ones, and those 3 are checked.
Success to me is a bad word, but they use the word correctly. The result was sent out, received and reported. Means NOTHING more.
I know the BOINC team is working on a way that the websites can see the bad results like this one, and start dropping their quota, and maybe even giving them a pop-up that their machine is giving bad results.
Yes, the team here can e-mail the person (if they have a valid e-mail account), and set their quota to 0 (as not to allow any more download). I hope that will be done.
(/edit)Looking further in this, I am wrong. It is affecting scores. Someone PLEASE stop this crazy machine, set their quota as 0. E-mail them. Thanks!
RE: Looking further in
)
I looked at several of the WUs; whether it affects the score or not depends on the timing of the other computers that return results. (This one is almost always first.) Because of the "throw out the high and the low" rule, this one is always thrown out as the low. The next two that return will set the credit, with the last one not being used. If that last one _would_ have been the "high", then this monster has caused a credit loss; the middle two are checked, the high claim of _those_ is thrown out, and the claim of the other is used. The difference between the claims of the middle two returns is the amount of the credit loss.
Example:
Host 1 (this one) - claims 0
Host 2 returns claiming 70
Host 3 returns claiming 60
Host 4 returns claiming 75
The validator runs as soon as Host 3's result is in. Host 1 and 2's claims are thrown out, and everybody gets 60 credits. Host 4 returns and automatically gets 60.
If Host 1 were not present, or returned "error" instead of "success", Host 3 and 4's claims would be thrown out and everybody would get 70.
If the order were different:
Host 1 (this one) - claims 0
Host 4 returns claiming 75
Host 2 returns claiming 70
Host 3 returns claiming 60
Validator runs after #2, throws out 1 and 4, everybody gets 70, no loss.
So, the quorum system is doing it's job, _reducing_ the impact of a bad host, on average. It can't totally eliminate the impact, so obviously it would be a "good thing"(tm) if this host could be strangled.
One of my machines (357407)
)
One of my machines (357407) is acting similar to this string, it claims credit, finishes with success, but is granted 0 credit, even though two other machines have responded. It shows the last 9 reports this way. Anyone know what's going on?
Ben, all your results from
)
Ben, all your results from that computer are invalid, but i can't tell why. The bytecount differs from the other hosts returning the same WU, as does the checksum. Basically, you're reporting "success", but the actual data you're sending back isn't even close to what the other computers are returning.
It obviously isn't extremely overclocked, but it could be a heat problem, or a RAM problem... though I'd expect more actual "errors" in those cases. Is that computer running any other projects? Any problems there? It's not _hurting_ anything here, other than maybe slowing down credit being granted for those WUs, but it's not doing you or the project any good, either.
Thanks, the machine is
)
Thanks, the machine is running einstein@home exclusively, and had been running OK for 4600 points. I've reloaded boinc.exe and will check the fan to see if it's too hot, but I don't think so. If that doesn't fix it I'll take it off line.
Shouldn't the validator send
)
Shouldn't the validator send the batch back where it came from after it recognised the invalid result(which it do) and not grant credit yet?
edit: and check everything to "Checked, but no consensus yet"
edit 2:maybe not.. Wiki about the validator.