Message boards :
Number crunching :
FIVE 'success' results & still pending..?
Message board moderation
Author | Message |
---|---|
UK_Nick Send message Joined: 4 Dec 00 Posts: 6 Credit: 15,555 RAC: 0 |
FIVE 'success' results & still pending..? <a href="http://www.nmvs.dsl.pipex.com/">Distributed Mania <img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=6&team=off&trans=off"></a> |
PT Send message Joined: 19 May 99 Posts: 231 Credit: 902,910 RAC: 0 |
Only five? ;-) I think I have close 30 or 40 pending... Happy crunching! Happy crunching |
SURVEYOR Send message Joined: 19 Oct 02 Posts: 375 Credit: 608,422 RAC: 0 |
It was sent out to the six computer, nothing will be granted until the validater has a chance to cross check the results. Peder check his link five results for the same wu, not five pending wu's Fred BOINC Alpha, BOINC Beta, LHC Alpha, Einstein Alpha |
Ingleside Send message Joined: 4 Feb 03 Posts: 1546 Credit: 15,832,022 RAC: 13 |
> It was sent out to the six computer, nothing will be granted until the > validater has a chance to cross check the results. The validator have already tried to validate this wu 3 times, not counting the time failed due to missing result-file, and for some reason haven't found a pair of results to be "similar enough". If it also fails the 4th validation-attempt when next result is reported, this wu error-outs and none gets any credit. |
UK_Nick Send message Joined: 4 Dec 00 Posts: 6 Credit: 15,555 RAC: 0 |
90% + of my 'success' results were still pending a few days ago but they've flowed in okay over the weekend. All five hosts marked 'success' have had many other results verified - strange that the same input data when crunched into output data is so different between five at least reasonably stable machines that not even 2 out of the 5 output data sets are close enough to verify as the same..? (My host 'Marion' only has a list of 6 results still showing, one of which is verified, but she must have had roughly 18 results verified succesfully to have 560 cobblestones to her name. My DC farm is mostly dedicated to Climate Prediction under BOINC at the moment but I still runch a few WUs for SETI-BOINC occasionally. <a href="http://www.nmvs.dsl.pipex.com/">Distributed Mania <img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=6&team=off&trans=off"></a> |
Keck_Komputers Send message Joined: 4 Jul 99 Posts: 1575 Credit: 4,152,111 RAC: 1 |
> 90% + of my 'success' results were still pending a few days ago but they've > flowed in okay over the weekend. <a> href="http://setiweb.ssl.berkeley.edu/workunit.php?wuid=1647046"> All five > hosts marked 'success' have had many other results verified[/url] - strange that > the same input data when crunched into output data is so different between > five at least reasonably stable machines that not even 2 out of the 5 > output data sets are close enough to verify as the same..? (My host 'Marion' > only has a list of 6 results still showing, one of which is verified, > but she must have had roughly 18 results verified succesfully to have > 560 cobblestones to her name. > > My DC farm is mostly > dedicated to Climate Prediction under BOINC at the moment but I still runch a > few WUs for SETI-BOINC occasionally. > > > One possible explanation is that no more than two of those 5 results were turned in by the same version. Unlikely since the reporting format is supposed to be the same in any different minor version as long as the major version is the same. It may also be a problem with the data. BOINC WIKI BOINCing since 2002/12/8 |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
My suspicion is that the earlier WU results from the original submitters may have been eaten by one of the various server problems...when the validater goes to try and compare results it finds 1 or 2 recent returned results but the older one(s), the file is missing, and out it goes again. No recent file server problems, so when 3 newer arrive on server maybe they get compared. Just a guess really. |
Stephen Balch Send message Joined: 20 Apr 00 Posts: 141 Credit: 13,912 RAC: 0 |
@Benher, I would think that a design error. If the validator attempts to find a result and cannot, for whatever reason, shouldn't it flag the record containing the status information (or code) displayed on the "Work Unit" page with a "Lost Record" status? It might also save processing time if the validator, or other processes, could interrogate the status of a result record by other than reading the table for the record, but I don't know how the tables are set up, or the time impact this may have. I have no idea how adding a new status to the system would percolate through the rest of the system, but I would suggest this is a change to be serously considered. Even if/when the current server problems are eliminated, or at least reduced considerably, there is always the chance of similar problems occurring again in the future if there is a hardware or database failure like those occurring in the fairly recent past. At the least, setting a "Lost Record" status in the "Work Unit" page data would greatly reduce the complaints I read about "old" results still having a "pending" status. Please pass this suggestion on as appropriate. Cheers, Stephen > My suspicion is that the earlier WU results from the original submitters may > have been eaten by one of the various server problems...when the validater > goes to try and compare results it finds 1 or 2 recent returned results but > the older one(s), the file is missing, and out it goes again. > > No recent file server problems, so when 3 newer arrive on server maybe they > get compared. > > Just a guess really. > <P>"I want to go dancing on the moon, I want to frolic in zero gravity!....", and now, I might be able to go someday! Thanks, SpaceShipOne and crew!<BR><a><img src="http://69.93.59.107/stats/banner.php?cpid=26cbd89db7fb85cbfe580729d76705c1"></a> |
Keck_Komputers Send message Joined: 4 Jul 99 Posts: 1575 Credit: 4,152,111 RAC: 1 |
@Stephen It may not be needed to add a lost record status indicator, at least from the projects point of view. BOINC will only send out so many copies of a workunit before the workunit is deemed unusable/corrupted at the source. I believe this defaults to 20 and am reasonably certain that is the number seti@home uses. From the users point of view this may not be as desirable since there will be 20 copies of the dead workunit before it is eliminated. In the case of missing files which seems to be the most common database problem now. If the input files are missing and unable to download that workunit should reach the 20 copies limit fairly quickly with minimal waste of users CPU time. If the results files are missing I would expect them to be regenerated as the resent workunits are returned. Thus the resent work would likely get validated but the originals would be deemed invalid since there is nothing to check them against. BOINC WIKI BOINCing since 2002/12/8 |
Ingleside Send message Joined: 4 Feb 03 Posts: 1546 Credit: 15,832,022 RAC: 13 |
> My suspicion is that the earlier WU results from the original submitters may > have been eaten by one of the various server problems...when the validater > goes to try and compare results it finds 1 or 2 recent returned results but > the older one(s), the file is missing, and out it goes again. The validator marks all missing "success"-results as "invalid", so this isn't the problem. > > No recent file server problems, so when 3 newer arrive on server maybe they > get compared. > The two results returned 5. & 8. December was compared against eachother, and against all the other 3 results, but no pair of results was "similar enough" so therefore re-issued once again. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.