ALMA Computing Report Number: COMP-70.55.00.00-001-A-REP

Scheduling Subsystem User Test 1 Report

Written by Debra Shepherd, based on an e-mail exchange between Mel
Wright (Scheduling subsystem scientist) and Allen Farris (subsystem
lead).  


12feb04


1. Goals:

   - Schedule simple simulated projects. 

   - Develop draft logs & evaluation output. 


2. New functionality tested:

   - Simulator version of scheduling code should be able to be
     installed at a users location and work offline with simulated
     SBs.

   - Draft version of simple heuristics for choosing SBs based on
     e.g. weather, HA-range desired, & priority should be available.

   - Log output should be available to see how the simulator code
     scheduled SBs.  


3. Summary: 

Mel Wright (subsystem scientist) successfully installed the scheduling
software in stand-alone simulator mode.  The software executed as
expected (except for one bug listed in Sect. 4).  Simple simulated
scheduling blocks were scheduled using heuristics defined in the
code.  The heuristics behaved as expected however a number of
improvements were identified to make them more appropriate for 'real'
projects.  Specific suggestions on how to improve graphical input and
output to the scheduler and on the types of plots needed to evaluate
the system status are presented in Sect. 5.  


4. Test issues: 

- I tried setting all weights = 0.0, except priority = 10.0
  This produced some NaN's

- The code tries to schedule at the highest elevation possible
  i.e. close to transit (LST = RA). This is good for best opacity and
  Tsys, but in many cases we need a range of LST on the same source to
  get good uvcoverage. This is more important for sparse arrays, but
  is also true for ALMA's longer configurations.


5. Suggested future improvements in the simulator: 

  - we need to generate some mechanism to get the required
    uv-coverage.

  - graphical input and output to show how the scheduler is
    performing. 

  - more sophistication in the scheduling algorithm itself.

- Suggested plot output: 

  1. The input list:
     projects listed vertically in priority order as horizontal bars,
     (color = priority ?)
     with their SB's plotted horizontally versus LST 0 to 24.
     i.e. Number of projects high x  24 hrs wide.

  2. The output schedule:
     days listed vertically as horizontal bars, with the scheduled
     SB's plotted horizontally versus LST 0 to 24. i.e. Number of
     scheduled days high x 24 hrs wide.  This looks like the schedules
     that observers are used to seeing, and could have daytime  as
     diagonal lines (a lighter shaded sunrise to sunset is more
     significant than the traditional midnight and 8am local time
     diagonal lines.) 

     [Other UT critical observation times like comets and solar system
     objects could also to shaded in if we want to get fancy.] 

     As the scheduling proceeds, the input plot is depopulated
     (de-colored ?) as the SBs are "executed", and the output schedule
     is filled in before your eyes.

  Both plots on the same page. We could eventually add another plot
  which shows the weather (opacity and seeing versus time) which is
  also filled in as LST and scheduling proceed. The opacity and seeing
  data come from the weather model in the simulation and from the real
  instruments in real life. The effect of the weather also shows up in
  the output schedule plot as the scheduling proceeds.