Minutes of Online Calibration Meeting 1/19/01

Minutes of Online Calibration Meeting 1/19/01

This meeting ended up being devoted primarily to
news items, partially due to low turnout among the
detector groups.

SERVER NEWS (Rick St. D.):
The online production database server was having hardware
problems over the holidays, but has been fixed. Unfortunately,
it did go down Friday morning after having been up with no
problems for 5 days, so investigations are still under way.

DATABASE TESTING (Rick St. D.):
A testing program has been developed which loads dummy
data into the calibration database based on the frequency
and channel count for each calibration. The frequency and
channel count are now REQUIRED when making java
definitions for use by the code generation. While running
the test program, the database can be monitored with a
nice visual tool SPOTLIGHT on Oracle. The testing
revealed two basic facts.
  1) The write time is linear with respect to the number
       of rows being written.
        - With the current table setup it takes 20 MINUTES
            TO WRITE A NORMAL SET OF CALIBRATIONS
            TO THE DATABASE.
        - the silicon group can gain a factor of 128 in write time
           by changing tables that write 1 row/strip to 1 row/chip.
        - the calorimeter and shower max groups can gain a
           factor of 32 or more in write time for their qie tables
           by going from 1 row/cap to 1 row/qie or other.
  2) The current sizing parameters for the tables in the
       database DO NOT ALLOW MORE THAN ONE DAY
       OF RUNNING BEFORE BLOWING OUT THE
       DATABASE.

SCHEMA CHANGES FOR CALIBRATION (Jack C.) :
The calibration database schema underwent large
changes. The database is composed of 3 sections
(better described in CDF5088 to be updated next
week): STORAGE, SETS, DEFINITIONS. Most
of the changes were to the DEFINITIONS sections.
The main points are:
   - there is a model for process_name's for valid sets
      based on calib_mode,destination,system.
   - the data_status field in calibrunlists is now
      constrained by a table with a list of possible
      data_status values.
   - the process_name field in valid_sets is now
      constrained by a table with a list of possible
      process_name values.
Bill B. brought up that some of the new tables
associated with process_name are beginning
to overlap with certain run configurations tables,
so the interface between the two databases
needs to be reviewed.

ONLINE VALID SETS (Jack C.):
The process for making and using valid sets
for calibrations in the online is in dire need of
development. A system for handling information
in the calorimeter crates is being prototyped
and was discussed at the meeting. The action
diagram is attached as an eps file. Green
boxes are implemented and tested, blue
boxes are prototyped, and red boxes are
vaporware. This diagram corresponds to a
use case or scenario. ALL OF THE DETECTOR
GROUPS NEED TO PROVIDE DESCRIPTIONS
OF THEIR USE CASES, so that the infrastructure
can be built.