WU sent to 20 clients. Just curious.
2 Dec 2016 14:36:09 UTC
Topic 203365
(moderation:
Greetings?
Just curious about:
https://einsteinathome.org/workunit/262276174
The results was "Too many results" (I I remember correctly.)
Most clients did not take more than one day - so it wasn't a timeout.
Not complaining - just curious..
Thanks, Jay
This happens if a bad beam is
)
This happens if a bad beam is not canceled. All of the successful results didn't validate because of too much noise in the beam. normally we catch those bad beams but lately there was no time to check them so they end themselves with this "too many results error.
Can the count be set lower
)
Can the count be set lower than 20 to reduce the invalids people get?
There is only one number for
)
There is only one number for all error types. Lowering it would potentially abort a workunit that is not noisy but produces a lot of host related error tasks (because tasks get send to hosts with local problems).
We are working on a fix to cancel noisy beams earlier and in a way that minimizes loss of compute power. In the meantime we also ran out of Arecibo beams to distribute. We don't have a continuous stream of data from Arecibo but get batches from time to time. That's why we focused on providing the GPU application for the FGRPB1 search to provide an alternative for BRP4G, this will shortly be made available for all EaH users.