0 credits for valid results


log in

Advanced search

Questions and Answers : Unix/Linux : 0 credits for valid results

Author Message
Profile Ananas
Volunteer tester
Send message
Joined: 14 Dec 01
Posts: 192
Credit: 2,319,430
RAC: 0
Germany
Message 882355 - Posted: 4 Apr 2009, 20:54:27 UTC

A friend posted this host in our forum today :

results.php?hostid=1198913 (not my host)

Those results are most likely valid, it's an incompatibility between CC 4.xx and the current BOINC API lib, that doesn't allow the application to report the runtime to the core client. The results are not affected.

I had the same effect at Einstein with CC4.19 on an old Coppermine 600 running Linux - Einstein does grant credits to such a machine though, ignoring the runtime of 0 seconds.

OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13614
Credit: 30,252,218
RAC: 19,969
United States
Message 882360 - Posted: 4 Apr 2009, 21:44:31 UTC

That is correct.

In the case of a validated quorum of two returned results, the lesser of the two claimed credits is granted. The reason why SETI does this is because they are open source software and anyone can edit the source code to make the science app claim high amounts of credit. Since SETI has a quorum of two and the lesser of the two is granted, this is sort of an anti-cheat measure.

The downside to this is older BOINC clients that do not accurately record their runtime to claim the correct amount of credits. Older BOINC clients used a benchmark * time method of claiming credit, but the credit system has changed after v4.49 to a FLOP-counting system which requires an accurate accounting of time to complete the workunit.

The only solution, if possible, is to upgrade the BOINC client to something newer than BOINC v4.49.
____________

Profile Ananas
Volunteer tester
Send message
Joined: 14 Dec 01
Posts: 192
Credit: 2,319,430
RAC: 0
Germany
Message 882408 - Posted: 5 Apr 2009, 0:14:27 UTC
Last modified: 5 Apr 2009, 0:23:32 UTC

Well, in case of old Linux core clients SETI should probably grant a low, but fixed minimum.

Someone has added some code lines to the API that seem to be thought to restore compatibility to older Linux core clients (using attach_shmem instead of attach_shmem_mmap) - but either it doesn't work or SETI uses an API version that doesn't have this code.


p.s.: On some boxes I couldn't use anything but 4.19 - trouble with the Squid authentication of all cURL BOINC clients back then.

OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13614
Credit: 30,252,218
RAC: 19,969
United States
Message 882414 - Posted: 5 Apr 2009, 0:23:21 UTC - in response to Message 882408.

Well, in case of old Linux core clients SETI should probably grant a low, but fixed minimum.


This has been asked many times by the people in the Number Crunching forum. After asking Project Administrator Dr. Eric Korpela personally, he has stated that he refuses to institute a minimum version because even if credits aren't granted, the science is still valid and useful, and he does not want to cut off any set-it-and-forget-it type users who don't even watch their credit or BOINC version.

Someone has added some code lines to the API that seem to be thought to restore compatibility to older Linux core clients (using attach_shmem instead of attach_shmem_mmap) - but either it doesn't work or SETI uses an API version that doesn't have this code.


Perhaps the user should work with a newer codeset of the API instead of using an older version, if possible. Otherwise, its a risk they take for using an older client that is no longer the recommended version.
____________

Profile Ananas
Volunteer tester
Send message
Joined: 14 Dec 01
Posts: 192
Credit: 2,319,430
RAC: 0
Germany
Message 882416 - Posted: 5 Apr 2009, 0:29:52 UTC
Last modified: 5 Apr 2009, 0:32:28 UTC

I am aware of all this - but people who use a current CC are affected too, if their partner uses an old core client.

My team mate complained about the validator in our forum, he assumed, that those 0 seconds results were invalid and the validator would still accept them and validate them against his valid ones.

His conclusion has been, that invalid results end up in the science database. Of course I cleared it up in our forum - but others might think the same when they see such a constellation without knowing the history.

OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13614
Credit: 30,252,218
RAC: 19,969
United States
Message 882430 - Posted: 5 Apr 2009, 0:54:00 UTC - in response to Message 882416.

I am aware of all this - but people who use a current CC are affected too, if their partner uses an old core client.


That's true, and the more users using a newer BOINC client there are, the less of a chance we have of getting paired up with one of these users, but the fact is we all have an equal chance of getting paired up, so at least its fair.

My team mate complained about the validator in our forum, he assumed, that those 0 seconds results were invalid and the validator would still accept them and validate them against his valid ones.

His conclusion has been, that invalid results end up in the science database. Of course I cleared it up in our forum - but others might think the same when they see such a constellation without knowing the history.


That would be why people who modify the code on their own should ask questions in the BOINC developers forum to get accurate answers to their assumptions, and not on individual team forums who would not have the correct answers.

____________

Profile Ageless
Avatar
Send message
Joined: 9 Jun 99
Posts: 12298
Credit: 2,591,785
RAC: 909
Netherlands
Message 882432 - Posted: 5 Apr 2009, 1:01:32 UTC - in response to Message 882430.

That would be why people who modify the code on their own should ask questions in the BOINC developers forum to get accurate answers to their assumptions, and not on individual team forums who would not have the correct answers.

Luckily Pineapple... um sorry, Ananas looks on both and is knowledgeable enough to answer them. :-)
____________
Jord

Fighting for the correct use of the apostrophe, together with Weird Al Yankovic

Pineapple
Send message
Joined: 22 Jun 03
Posts: 1
Credit: 10,887
RAC: 0
Germany
Message 882461 - Posted: 5 Apr 2009, 5:36:29 UTC - in response to Message 882432.

Luckily Pineapple... um sorry, Ananas looks on both and is knowledgeable enough to answer them. :-)


Well ... actually this is me too :-)

Questions and Answers : Unix/Linux : 0 credits for valid results

Copyright © 2014 University of California