Change where BOINC stores unprocessed WUs?


log in

Advanced search

Questions and Answers : Wish list : Change where BOINC stores unprocessed WUs?

Author Message
Profile K1VRA
Send message
Joined: 28 May 99
Posts: 4
Credit: 12,798,581
RAC: 5,672
United States
Message 619538 - Posted: 15 Aug 2007, 12:50:23 UTC

I have several, networked PCs, each is using BOINC for s@h and each has it's own cache of WUs waiting to be processed. If every PC could 'look' in the same place (say, a shared network drive) for it's next WU, the local cache of WUs waiting to be processed would be much smaller. How about letting me choose where the WUs are stored?
____________

Profile Pooh Bear 27
Volunteer tester
Avatar
Send message
Joined: 14 Jul 03
Posts: 3221
Credit: 2,214,164
RAC: 9,361
United States
Message 619672 - Posted: 15 Aug 2007, 17:59:17 UTC

BOINC was created to make sure cheating doesn't happen. So, they limited the ability to do server caching as you have recommended. Each machine that downloads a unit should be the same one that returns the unit. This way they can keep a better integrity on the data.


____________

Profile K1VRA
Send message
Joined: 28 May 99
Posts: 4
Credit: 12,798,581
RAC: 5,672
United States
Message 621052 - Posted: 17 Aug 2007, 14:25:51 UTC - in response to Message 619672.
Last modified: 17 Aug 2007, 14:39:48 UTC

BOINC was created to make sure cheating doesn't happen. So, they limited the ability to do server caching as you have recommended. Each machine that downloads a unit should be the same one that returns the unit. This way they can keep a better integrity on the data.



Thanks for your response. So I take it that, even though the WUs were all downloaded by the same individual (not a 'team') user, albeit processed on different PCs, that would not be sufficient for tracking? (I do understand the need for tracking and maintaining the integrity of the research.)

imho, is there really much difference between my PCs taking WUs off SETI's server or taking them from my own 'intermediate' server? The machine that actually processed the WU would be the one reporting it back. I just saw an opportunity to streamline the process while reducing the total number of WUs stored here (which makes them un-available for others to process - each of my machines caches several WUs ahead). Say, in order to keep my machines busy in the event of a SETI outage, or my own ISP's failure.

Thanks, again.
____________

John McLeod VII
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 15 Jul 99
Posts: 24376
Credit: 519,750
RAC: 37
United States
Message 621499 - Posted: 18 Aug 2007, 0:05:53 UTC - in response to Message 621052.

BOINC was created to make sure cheating doesn't happen. So, they limited the ability to do server caching as you have recommended. Each machine that downloads a unit should be the same one that returns the unit. This way they can keep a better integrity on the data.



Thanks for your response. So I take it that, even though the WUs were all downloaded by the same individual (not a 'team') user, albeit processed on different PCs, that would not be sufficient for tracking? (I do understand the need for tracking and maintaining the integrity of the research.)

imho, is there really much difference between my PCs taking WUs off SETI's server or taking them from my own 'intermediate' server? The machine that actually processed the WU would be the one reporting it back. I just saw an opportunity to streamline the process while reducing the total number of WUs stored here (which makes them un-available for others to process - each of my machines caches several WUs ahead). Say, in order to keep my machines busy in the event of a SETI outage, or my own ISP's failure.

Thanks, again.

The problem is that some users were abusing this ability in Classic to get credit for doing a task multiple times by copying the completed task to several computers. Some BOINC projects with very few users actually allow two Tasks from the same Work Unit to be assigned to different computers of the same user, while this not the preferred method, it can be the acceptable. But assigning the the two Tasks from the from the same WU to the same computer essentially removes the cross check completely.
____________


BOINC WIKI

Profile Kitter
Send message
Joined: 19 Aug 03
Posts: 5
Credit: 6,353,643
RAC: 1,401
Czech Republic
Message 622259 - Posted: 18 Aug 2007, 21:57:51 UTC
Last modified: 18 Aug 2007, 21:59:12 UTC

Exactly, each machine that downloads a unit should be the same one that returns the unit. But I have a trick for this. You can run a new job at boinc, then pause it, remember its name, copy state.sah and init_file.xml from the corresponding slot folder, corresponding result and workunit file from the project folder to some new folder and setiathome_etcetc.exe(better one is optimized (eg. SSE2) version from lunatics.at which does not require any extra files) too. Then just rename workunit and result file to work_unit.sah and result one to result.sah. Run the exe file and wait till it completes(eg. at some other computer over network). After the completition do exactly reversed process. Unpause the job and it should resume somewhere around 99%. This process should be sensitive to seti servers when made correctly. So remember to be care - full. Oh and when copying the files back, better shut boinc down, in the first stage it souldnt be necessary.
____________

Questions and Answers : Wish list : Change where BOINC stores unprocessed WUs?

Copyright © 2014 University of California