Joe already proposed some method BOINC could implement to guard project (SETI for now, maybe others will added later) in such situation: do quota deduction when task was finished much faster than "usual". If task will validated later, quote will be restored (validation bonus) but it allows to do fast inhibition of broken host. As I already said, time scale for such deviation is hours, not days or weeks (so just inhibition by invalid results not apply here), but such preventive quota supperessing can help. And BOINC doesn't need to know details of results returned, all that it needs is mean completion time and particular completion time.
----- Original Message ----- From: "David Anderson" <[email protected]> To: "Raistmer" <[email protected]> Cc: "Eric Korpela" <[email protected]>; <[email protected]> Sent: Friday, June 04, 2010 12:59 AM Subject: Re: [boinc_dev] host punishment mechanism revisited >I don't think BOINC can or should solve this problem; > it's a problem with the s...@home CUDA app. > Eric, are you aware of this? > -- David > > Raistmer wrote: >> One more example when 2 overflows suppressed single non-overflowed (and >> much probably valid) result and passed through Validator into database. >> These CUDA-induced overflows can pollute SETI's database. >> http://setiathome.berkeley.edu/workunit.php?wuid=618018953 > _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
