question about last checkpoint
29 Jun 2014 5:17:57 UTC
Topic 197618
(moderation:
Hello.
I used only gravitational wave S6... 1.05 but my WU did not saved any progress. The WU is at 6.5% (after 27min work) but there is no cpu last checkpoint. And if i shut down boinc client or my pc, the data is lost and this WU start again from 0,00%...... What can i do?
question about last checkpoint
)
The Gravitational wave search writes checkpoints quite infrequently and might need anything from 30 min to more than 1 hour between them.
One of my currently running S6 tasks took 44 min between checkpoints. That's on a i7 3770K, slower processors will require even more time between checkpoints.
The developers are aware of the problem and have stated that it's difficult to change the frequency of checkpoints for this search, hopefully that can be improved on the next search.
If this is unacceptable for you I recommend that you go to your Einstein@home prefs and opt out of "Gravitational Wave S6 Directed Search (CasA)".
The last batch I saw like
)
The last batch I saw like that checkpointed 11 times during the run, or after each 9.09% of progress.
With the 'recommended' 7.2.42 version of BOINC that you are using, the early stages of progress (before the first checkpoint) are not accurately recorded: they are simply an estimate based on time spent so far (earlier versions of BOINC displayed no progress at all during this phase, which was even worse).
If you can, leave the computer running until the progress% shows exactly 9.090% (it should then stop changing, until a second big jump to 18.181%). You should then be able to look up the CPU time at the first checkpoint: you would need to leave the computer running for at least that long in each session, to make progress with these tasks. If that's not possible, you might be better off choosing a different type of Einstein or BOINC work.
It seems to me that the
)
It seems to me that the gamma-ray pulsar search isn't very good in this regard either. It would be nice to be able to see the time of the last checkpoint from the BOINC Manager interface.
The big size and relatively tight deadlines of the Einstein work units mean that this project is not ideal for every BOINC user. Fast machines that are always on don't really have the checkpoint problem.