[Clas12_first_exp] Analysis trains and skims

[Raffaella De Vita]

Dear All,
here is more information about the status of analysis trains and about 
the feedback that would be useful to have from collaborators to start 
more accurate tests.

Gagik has released the first version of the analysis train software.
- Analysis trains are conceived to do "event selection": they use DSTs 
as input, i.e. cooked files with all events but only the banks that are 
relevant for the analysis (see Nathan's documentation at 
https://clasweb.jlab.org/wiki/index.php/CLAS12_DSTs), and for now to 
output skims for different event topologies with the same hipo format 
and the same banks of the input files.
- Different skims, defined based on user requests, make the train 
"wagons". After collecting user requests for wagons, trains will be run 
centrally by a single chef, to optimize the efficiency of the skimming 
process and avoid I/O problems that are expected to arise with several 
people trying to access large DST volumes simultaneously.
- Skim conditions can be defined based on the number and type of final 
state particles as explained in Gagik's documentation at 
https://userweb.jlab.org/~gavalian/docs/sphinx/hipo/html/chapters/analysis_train.html#event-filters. 


To define the first trains and start testing the overall process, we 
would like to ask collaborators to come up with requests for final state 
topologies for their favorite reactions:
- check the instructions at 
https://userweb.jlab.org/~gavalian/docs/sphinx/hipo/html/chapters/analysis_train.html#event-filters 
and send us your "wagon" definition;
- keep in mind that, for a skim to be useful, it should correspond to a 
small fraction of the input data; you can estimate the reduction factor 
of your proposed skimming condition using the sample DSTs available in 
/work/clas12/clas12/data/calib/dst_5bp6p1/; the overall train output 
size (sum of all skims that will be collected) should be less than 50% 
of the input DST;
- once requests will be collected and reviewed to remove possible 
duplications, a train will be assembled and tests will begin.

Best regards,
     Raffaella





Latifa Elouadrhiri wrote:
>
> Dear all,
>
> The meeting this morning is cancelled.
>
> The meeting agenda for next week will focus on the following:
>
> *Spring Run*
>
>   * Status of the data processing
>   * Analysis trains for each one of the major analysis group: please
>     send me the requested filtering so far I have the one from SIDIS
>     (Harut) and From Nucleon Structure (Ralf).
>
> *Fall Run*
>
>   * We will review where we are in term of tracking efficiency based
>     on the latest release and make decision on the luminosity for the
>     upcoming run
>   * List of run coordinators
>   * Run plan
>
> Best regards,
>
> Latifa
>
>
>
>
> _______________________________________________
> Clas12_first_exp mailing list
> Clas12_first_exp at jlab.org
> https://mailman.jlab.org/mailman/listinfo/clas12_first_exp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/clas12_first_exp/attachments/20180801/a939cfe5/attachment.html>