THE QUEUE SCHEDULING PROGRAM WITH CFH12k:

Guidelines for the Telescope Allocation Committees


Pierre Martin, Dennis Crabtree
April 1999

Background

This document presents some guidelines for the Telescope Allocation Committees (TAC) in their evaluation of the proposals submitted for the queue scheduling/service observing (QS/SO) program with CFH12k. Following a recommendation by the Scientific Advisory Committee, it is planned that a very large fraction (if not all) CFH12k observations for the semester 1999II will be executed in a QS/SO mode. From the queue experiments done at the NTT and WIYN telescopes, the major advantages and problems of the QS/SO mode have now been identified. Specific tasks must be assigned to the TACs in their evaluation of the proposals for the QS/SO mode to work properly. Below, some basic execution requirements for the QS/SO mode with CFH12k are described. In the second section, the specific tasks requested from the TACs regarding the CFH12k proposals are detailed. Other issues, among them some TAC actions concerning classical mode observations, are discussed in the third section.

For more information concerning the queued program with CFH12k, the members of the TACs can also consult three other documents at the CFHT Web site here.

A. Queue Scheduling: Basic Execution Requirements

The strategy of the QS/SO mode permits to gain in telescope efficiency and science productivity. The advantages of the QS/SO mode have been previously extensively documented:

• QS allows an optimum match of the program requirements to the current atmospheric conditions. The most impressive gains are expected for programs requiring excellent (often exceptional) conditions.
• The QS mode allows the execution of a larger fraction of approved programs.
• QS optimizes the distribution of program according to varying, predictable conditions (e.g. dark vs. gray time).
• Programs with unusual constraints (e.g. target-of-opportunity, monitoring, observations for only a fraction of a night) can be more easily integrated in a QS/SO scheme.

These goals can be achieved by defining a set of rules controlling the execution of the queued programs. However, following the results of the NTT and WIYN queue experiments, it is now clear that the efficiency of the QS/SO mode depends strongly on how the queue is filled. It is extremely important that the queue is filled with programs requiring a wide range of conditions. For instance, it is particularly critical to not approve only programs requesting dark time since it will become extremely hard to execute these programs in the queue within the amount of dark time available. This is the key point for the TACs: By considering realistic conditions (i.e. number of photometric nights, seeing statistics, dark time) and the number of nights allocated for the QS/SO mode, the proposals accepted should not only include programs requiring the best conditions but also other proposals with less severe constraints. As a reference, the following table displays some rough statistics on weather conditions on Mauna Kea (observations done with FOCAM at CFHT between 1993 and 1995):

Table 1: Conditions of Mauna Kea

$$\pm$$

$$\leq$$ 5%

$$\leq$$ 25%

$$\leq$$ 30%

$$\leq$$ 25%

$$\leq$$ 15%

IQ > 1.2'' 5%

$$\sim$$ Usable Nights

$$\sim$$ Lost to weather

$$\sim$$ Usable Photometric Nights

The selection of programs will be executed automatically at CFHT with the scheduling software and reviewed by the observers. This task will be done following a weight system based on the program specifications, the actual conditions, and evaluation from the TAC. During the evaluation process, the TAC must provide the following information (detailed in the next section):

1.

A qualitative ranking for each proposal (``must do'', ``priority'', ``best effort'')

2.

A quantitative ranking for the ``priority'' and ``best effort'' lists

3.

The maximum total integration time (I-time) allocated for each program.

4.

The total time to be allocated for the QS/SO mode for the relevant semester with respect to the total I-time allocated.

5.

Specific conditions for observations to be conducted in classical mode, if requested.

It is also important here to briefly describe the policy that we propose to follow for carrying out the queued programs. The ultimate goal of QS/SO is, of course, to be able to complete the largest number of programs as possible within specifications. We propose to adopt the following rules, based on the queue experiment at the WIYN telescope, concerning the degree of completion of the programs:

• Programs started and at least 75% completed (but not fully completed), during the current semester may be completed during the next contiguous semester at the discretion of the CFHT Director. Otherwise, all uncompleted programs will be terminated at the end of each semester and will not be re-started until they are again reviewed by TAC.
• Programs not started by the end of the current semester will not be scheduled in any subsequent semester unless the proposal is re-submitted to the TAC.

We believe that this policy will result in a larger number of highly-ranked proposals to be completed since these proposals will have moce chance to reach the minimum degree of completion of 75% during the current semester.

B. CFH12k: Proposal Evaluation

B.1 Qualitative Ranking

As stated above, the QS/SO mode can only work efficiently if the proposals in the queue cover a wide range of observing conditions. The approved proposals should not exclusively request the best seeing conditions but other proposals with less severe contraints must also be considered. This does not mean that the latter will be executed, especially if their science merit is much lower, but if the conditions are completely inadequate for the most restricting programs, they will provide a satisfactory backup to avoid any waste in telescope time.

As a first step in the TAC evaluation, we propose that all the proposals be divided into four categories:

• Class A. This is the ``must-do'' programs. These programs will always have the highest priority in the queue list and will be executed unless the conditions are not adequate or have deteriorated. These programs will have the highest absolute quantitative ranking. No more than 15% of the total integration time recommended for the QS/SO should be in this class. The QS/SO mode aims to have a level of completion of 100% for these programs.
• Class B. This is the ``priority'' list and constitutes the bulk of the queue database. About 60-75% of all the proposals approved should be included in this class. We aim to have a level of completion between 75% and 100% for these programs, with the highest ranked proposals having a higher probability of completion.
• Class C. This is the ``best effort'' list of programs. These programs deserve some observing time but will not be executed during the best observing conditions. About 10-20% of all the approved proposals should be recommended for this class. A fraction of these programs might be fully performed but most likely, some programs will be completed at 50-75% and less while other ones will not be started.
• Class D. This last class includes rejected proposals. These programs do not receive any telescope time and are not included in the QS/SO database.

B.2 Quantitative Ranking

An important element in the selection of executable programs by the scheduling software is the TAC ranking. Although the science grade does not have the highest weight in the decision tree leading to the selection of the programs (actual sky conditions are), it becomes more important when one program has to be chosen among a set of programs requiring the same conditions.

All the programs in Class A will receive the same absolute quantitative priority. However, the programs classified as B or C will receive an absolute rank according to the TAC evaluation for these groups. So, TAC should rank all the approved B and C proposals as in the past.

Following this evaluation, this quantitative ranking will be apply in the queue database by the CFHT QS/SO team:

Table 2: Proposal Ranking

Class A No Ranking 1.00

Class B First 20% 1.25

$$\leq$$ 1.50

$$\leq$$ 1.75

$$\leq$$ 2.00

$$\leq$$ 2.25

Class C First 30% 2.50

$$\leq$$ 2.75

$$\leq$$ 3.00

B.3 Integration Time

The time given for a specific program is accounted as the integration time (I-time), that is, the time at the telescope under a clear sky with the camera shutter open and under conditions as requested in the proposal. This does NOT include overheads associated with the telescope pointing and camera readout time. Calibrations, unless specifically required by the programs, will be conducted through the QS/SO run and should not be included in the I-time. In the new electronic Phase I form, the total I-time requested can be found on the first page of the proposal and is justified in the list of targets and the technical feasibility section. However, for the queue database, TAC must give the approved I-time (in hours, not nights!) for each program. This value is automatically entered in the proposal database and is used during the Phase II proposal submission by the PIs to define their observations. During the execution of the queue, the I-time is used as a reference to evaluate the fraction of the program that has been completed until fully done.

B.4 QS/SO Allocated Time

Another task requested from TAC is the allocation of time for the QS/SO program. This is, of course, strongly related to the integration time described above. However, since the I-time does not cover the overheads, the total efficiency of the observation process must be considered.

Following our experience with CFH12k during the beginning of the semester 1999I, we are expecting a total observing efficiency of about 80% with CFH12k queued programs. The remaining time will be mostly spent on the field acquisition, readout and the calibration plan (program with the highest priority during photometric nights). Thus, we can calculate the number of nights that should be allocated for all the queued programs according to this efficiency and the statistics on weather conditions for Mauna Kea (e.g. Table 1). Roughly, about 100 nights are required if the total I-time requested is 650 hours. Again, however, it is preferable to overfill the queue to make sure that enough programs are available to cover all the time allocated for QS/SO.

C. Other Issues

C.1 Classical Mode

Some specific programs (e.g. Hippo polarimetry) are not suitable for QS/SO. If the investigators think that their program should not be executed in a QS/SO mode, they should have justified their request in their proposals. This justification should be evaluated by the TAC and if it appears that the observations should be carried out in a classical mode, TAC should be very specific on the sky brightness (dark time, gray, bright) under which these programs should be performed. We will build our schedule to accommodate these requirements and the selected dates will be fixed (that is, no switch allowed with queue nights). The other queued programs will be executed around these dates. We recommend, however, that TAC minimizes the number of nights allowed in a classical mode, if possible. Only exceptional circumstances or programs that cannot be executed in a QS/SO scheme can justify the classical mode.

C.2 Discretionary Time

A certain number of hours of discretionary time (DT) will be available for the CFHT astronomers. This time should not be included in the TAC calculation of the time allocated for the QS/SO observations. The Senior Resident Astronomer will review the CFHT astronomer requests and add the appropriate number of hours required to conduct these programs.

C.3 Time Accounting

It has been a concern that accounting of time for the different partners in QS/SO will be difficult and could result in some imbalance. It is probably true on the short term but inside a semester, and more so for a full year, the integration time executed will be very close to the actual contributions of the CFHT partners. With the CFHT database, it will be very easy to follow exactly how much time is spent for a given partner across a semester. Regularly, the queue coordinator will review the queued programs done and those available in the database. If some actions must be done to keep the carried out I-time close to the relative contribution of the partners, specific instructions will be given to the observers for the execution of the queue. However, such interventions are considered exceptional and will only be applied in case of obvious discrepancy between the different agencies. At this time, no TAC actions are necessary for time accounting of each Agency.

D. Summary

The CFH12k QS/SO mode requests a little more effort for the TAC members in their evaluation of the Phase I proposals. The main differences are found in the the ranking procedure, the calculation of the I-time for each individual programs, and the allocation of time for the QS/SO program. Since the semester 1999II really represents the first attempt from CFHT to fully implement this mode, we strongly recommend that the TACs overfill the queue database. This will allow more flexibility during the observations. This strategy might result in several programs not completed or even started but will avoid to have the telescope standing still under sky conditions good enough for astronomical observations.