What's in a Production Request

If this is a new kind of processing for the experiment, the production group should not be the first ones to make the files you want. You should already have in-hand examples of the files you want us to produce a lot of. Based on running that example and examining the output of TimeTracker and MemoryTracker at the end of the job you should be able to answer the following questions:

Processing an existing file (e.g. reco/pid on data or MC)

  • What are the input files?
    • SAM dataset(s) would be ideal, but a specific description will work to (nd/fd, data/mc, release, special conditions, etc.)
    • How many are there? How many events to do they typically have/file?
  • What release do you need to run in?
    • Better if it is a tagged release, but we can work with a snapshot.
  • What fhicl file is being run?
    • Is that fhicl file already committed to novaproduction? If not, please do that now and ask us to make a new release of that product.
  • How long does it take to process one event?
  • What is the peak memory needed for the job?
  • How soon do you need the files?

Generating Simulation

  • Can make_sim_fcl produce the simulation files you need?
    • If so, what's the right configuration for producing those fhicl files?
    • This is where things like ND/FD, gain, run matching, etc. are all specified.
    • If not, it needs to be modified to do what you need.
  • Do you need overlay?
    • If so, are the necessary scripts committed in NovaGridUtils?
    • How are they configured?
  • How long does it take to produce 1 event on average?
  • How many events total do you need?
  • What is the peak memory needed for the job?
  • How soon do you need the files?