On 13.12.2016 17:01, Adrian Worley wrote:
> Good day,
>
> I have discovered a problem at the pogs project, the discussion of the 
> problem seems to imply that the problem is a BOINC problem, not a pogs 
> problem, so I feel it correct to ask here.
> I found the problem whilst checking some result times. It appeared that my 
> very fast machine was being seriously outperformed by some much older 
> systems. They have done some investigating and found some very odd figures in 
> their database. My machine returned a work unit in 6:47 my wingman in 2:09 - 
> yet taking the recorded start and end times, these figures cannot be correct,
> My machine    05:07:18 to 07:22:03
> My wingman  01:04:17 to 05:23:31
> Suggestions have been made suggesting problems with the way BOINC calculates 
> these values, however, I have not seen this kind of thing at any of the other 
> 15 projects my machine is connected too. Are there any known faults or issues 
> within BOINC that could /sometimes/ account for these reults?
> Med venlig hilsen,
> Adrian Worley.

Here is some more information:

Original discussion on POGS:
https://pogs.theskynet.org/pogs/forum_thread.php?id=758

Wrapper output of the task with high runtime:
http://pogs.theskynet.org/pogs/result.php?resultid=77706342 or
http://pogs.theskynet.org/pogs/result.php?resultid=78451868

The task seems to be a bundle of 21 separate calls to a legacy
application. The wrapper outputs the CPU time of each call. In this
example the last of those calls is reported to have taken 376,761 sec
CPU time when according to the timestamps it only took about 6 minutes
real time. The overall elapsed time is reported as 24,442 sec and
overall cpu time is 195,536 sec.

> 07:11:14 (7772): wrapper: running ./fit_sed (20 filters.dat observations.dat)
> 07:16:38 (7772): ./fit_sed exited; CPU time 319.984375
> 07:16:38 (7772): wrapper: running ./fit_sed (21 filters.dat observations.dat)
> 07:22:02 (7772): ./fit_sed exited; CPU time 376761.640625
> 07:22:02 (7772): wrapper: running ./concat (21 output.fit)
> 07:22:03 (7772): ./concat exited; CPU time 0.000000
> 07:22:03 (7772): called boinc_finish(0)
This is client 7.6.22/7.6.33 one Windows 8.1 Core using wrapper 7.5.26014.

There is also an example where the miscalculated cpu time is not the
last call: http://pogs.theskynet.org/pogs/result.php?resultid=78432167

It would be nice to know if this happens only sometimes or just in
combination with other things. The POGS Admins should be able to gather
some data from their DB and find correlations. I'm not sure if they read
this list so someone may need to convey this information.

There is a newer wrapper available but as far as I can see none of the
commits mention cpu time fixes.

Regards
Christian

_______________________________________________
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.

Reply via email to