Snapshot Scheduling A

advertisement
Internal Technical Memorandum ITM-1998-02
Snapshot Scheduling
Merle Reinhart, Anuradha Koratkar, and Jim Younger
October 1, 1998
Concurrence:
ABSTRACT
This report describes the procedures used to select Snapshot exposures for execution. When Snapshot visits are ready to be placed on a flight calendar, they are added to the snapshot Pool automatically. The method for selecting a visit in the Pool is an “ordered” random sample of candidates
within six categories ordered by Cycle number (higher to lower) and percentage of visits that have
been scheduled or completed. A proposal that has less than 50% of its visits scheduled or completed is given preference within a cycle.
1. Introduction
Prior to July, 1998, a Snapshot (SNAP) visit was added manually to the SNAP Pool when a Program Coordinator (PC) told Science Planning and Scheduling that it was “ready for scheduling.”
Once a visit was placed in the SNAP pool, its scheduling priority was determined by the end of its
assigned scheduling window (earlier end times had higher priority) and the Proposal/visit-id number (lower numbers had higher priority). Due to the large number of visits in the SNAP Pool, and
the biases inherent in the old selection method, some proposals achieved a high percentage of
completion and others languished at a low percentage. As of July, 1998, SNAPs are placed in the
pool automatically, and a revised set of rules for selecting SNAPs for scheduling has been implemented. These changes should result in a more random sampling of the SNAP Pool.
2. Procedure
Once a SNAP visit status is set to “scheduling” in the planning and scheduling database by a PC,
the visit is automatically a candidate in the SNAP Pool.
2.1 Assembling a SNAP Candidate List
When a flight calendar is ready for the addition of SNAPs, the calendar supervisor queries the
database for a list of SNAP candidates (usually 200) from the pool of visits that have schedulability windows overlapping the flight calendar and are available for scheduling. This candidate list is
constructed in three steps. First, the query assembles a list of proposals still active (i.e., that do not
have a status of failed, completed, or withdrawn) in the three most recent cycles. Then, the visits
are divided into two groups for each proposal: one having a visit status of ready for scheduling and
the other having a status of complete or scheduled on an existing calendar. Based on these group-
1
Internal Technical Memorandum ITM-1998-02
ings, a scheduling completion percentage of a proposal is computed. Finally, six categories are
constructed (two for each Cycle) based upon cycle number and completion percentage. The first
category within a Cycle has a completion percentage less than a 50%; the second has a completion
percentage equal to or greater than 50%. The candidates in each of these six categories are then
independently randomized.
The visit candidates for a flight calendar are selected from the six categories in the following order
until the requested number of SNAP visits (usually 200) have been selected:
1. Cycle N proposals with a completion percentage <50%.
2. Cycle N proposals with a completion percentage ≥50%.
3. Cycle N–1 proposals with a completion percentage <50%.
4. Cycle N–1 proposals with a completion percentage ≥50%.
5. Cycle N–2 proposals with a completion percentage <50%.
6. Cycle N–2 proposals with a completion percentage ≥50%.
Here N is the current cycle number.
The situation where SNAP visits are selected from a total of three different Cycles will probably
occur in the 2 to 3 month overlap of a current Cycle and the nominal start of the next Cycle when
the next Cycle’s proposals are being prepared for flight readiness. When the next cycle nominal
start time is reached, the N–2 cycle’s SNAP visits will be set to “withdrawn.” Consequently, the
Cycle N–2 categories (categories 5 and 6) are usually empty. This three-Cycle situation is necessary because there is no way to determine with a database query the location of the nominal cycle
boundaries.
2.2 Adding SNAP Candidates to a Flight Calendar
The list of 200 candidates is reassembled in order of decreasing alignment time (largest to smallest), and the available gaps in a calendar are assembled in order of increasing gap size (smallest to
largest). The first candidate (with the longest alignment time) is checked against the list of gaps,
starting with the smallest gap, and is placed into the first gap that fits, if one is found. Then the second candidate in the list is similarly checked against the gaps and placed on the calendar if
possible.This procedure is continued down through the entire list of candidates. On average, 10
visits find a place on a typical flight calendar.
3. Flight Preparation Requirements
PCs who prepare SNAPs for scheduling should select a start window corresponding to the nominal
start time of the cycle they belong to and an end window corresponding to the nominal end time of
the next cycle (i.e., they should get prepped for a two year period). This should minimize the
amount of rework of SNAP proposals as well as the effect of the aforementioned three-cycle
situation.
2
Download