Source-linked AI summary
Ten Year Review of Queue Scheduling of the Hobby-Eberly Telescope
Matthew Shetrone, Mark E. Cornell, James R. Fowler, Niall Gaffney, Benjamin Laws, Jeff Mader, Cloud Mason, Stephen Odewahn, Brian Roman, Sergey Rostopchin, Donald P. Schneider, James Umbarger, Amy Westfall
TL;DR
The HET’s narrow observing windows and complex, multi-institution program mix make queue scheduling a central operational challenge. The paper reviews a decade of queue methods and their evolution, finding that HET operations have been successful for some program types while identifying limits in staff knowledge documentation and partner access to observing conditions.
Problem
HET’s narrow observing windows and diverse institutional programs make scheduling science, targets of opportunity, and engineering tasks across changing conditions complex.
Source: §1 · pp. 3–4Method
The paper reviews a decade of HET queue operations and lessons learned, including an evolving, customer-oriented process that involves PIs in operations.
Source: §1 · p. 3, §5 · p. 14Results
HET operations were most successful with target-of-opportunity and synoptic programs and surveys spread widely across the sky; most incomplete programs were pencil-beam surveys, and most failures stemmed from technical problems.
Source: §5 · p. 14Takeaways & Limitations
A customer-oriented queue can be successful when software and observing practices evolve with PI and TAC needs and PIs participate actively.
Source: §5 · p. 14The operations team lacked a living repository of telescope and instrument performance knowledge, a gap the authors say could be addressed with one additional FTE.
Source: §5 · p. 16
Abstract
from arXiv · showhide
This paper presents a summary of the first 10 years of operating the Hobby-Eberly Telescope (HET) in queue mode. The scheduling can be quite complex but has worked effectively for obtaining the most science possible with this uniquely designed telescope. The queue must handle dozens of separate scientific programs, the involvement of a number of institutions with individual Telescope Allocation Committees as well as engineering and instrument commissioning. We have continuously revised our queue operations as we have learned from experience. The flexibility of the queue and the simultaneous availability of three instruments, along with a staff trained for all aspects of telescope and instrumentation operation, have allowed optimum use to be made of variable weather conditions and have proven to be especially effective at accommodating targets of opportunity and engineering tasks. In this paper we review the methodology of the HET queue along with its strengths and weaknesses.
1. Introduction
The HET’s narrow observing windows make dedicated-night scheduling difficult, while queue operations can match varied programs to changing conditions. The paper reviews a decade of HET queue operations, including its institutional constraints, advantages, and costs.
1. Introduction
- Queue scheduling lets an observatory team select among programs as conditions and priorities change, with the promise of increased science productivity.It can match conditions to program constraints, support time-critical observations and targets of opportunity, and use experienced staff to take data.Source: §1 · p. 2
- Queue operations can cost more to staff than dedicated observing and raise concerns about data quality, discovery opportunities, program control, and observer creativity.The paper reviews a decade of operating the HET queue and the lessons learned.Source: §1 · pp. 2–3
- The HET combines a fixed-altitude, narrow-field telescope design with three facility spectrographs, shaping the operational setting for queue observing.Its observable sky forms a ring centered on the zenith, with targets available for limited intervals.Source: §1 · p. 3
- Because HET targets have only one or two narrow observing windows per night, tightly packed dedicated schedules are vulnerable to conflicts and disruptions.A single 7-hour target could take more than three HET nights, and standards, competing priorities, and glitches complicate scheduling.Source: §1 · pp. 3–4
- HET queue operations must balance time shares, including dark and light time, across five partner institutions with separate allocation processes.Queue scheduling was built into HET’s design to support temporal projects, targets of opportunity, and surveys.Source: §1 · p. 4
2. Development of HET Queue Scheduling
HET queue operations developed around independent institutional allocations, a shared target database, and resident astronomers who planned and updated observations. The process also had to resolve conflicts among programs and accommodate different proposal cycles.
2. Development of HET Queue Scheduling
- The initial staffing plan reserved 85% of nights for queue observing and assigned resident astronomers to select targets, configure instruments, execute observations, and monitor data quality.Telescope operators handled telescope movement, guiding, and weather monitoring; dedicated observing was not requested, and the facility manager’s planned RA duties proved unfeasible.Source: §2 · p. 4
- Independent Telescope Allocation Committees assigned hours and priorities, while each program submitted target, exposure, instrument, and observing-condition requirements for execution.Priority 1 was intended for targets of opportunity or especially time-critical observations.Source: §2 · pp. 5–6
- The partners used trimester proposal cycles, whereas NOAO retained unobserved priority 1 and 2 targets across semesters, trading PI engagement against simpler completion of high-priority programs.Most partner proposals were removed each cycle, with some long-term programs requiring new target lists.Source: §2 · p. 5
- Independent TAC decisions produced conflicts, including duplicate targets, competing time-critical requests, and gaps in available targets, which resident astronomers generally resolved with the TACs.A few cases required a decision by the HET Scientist, whose charge was to maximize telescope scientific productivity.Source: §2 · p. 6
- A shared target database fed htopx, which displayed currently available targets and ranked bright- and dark-time targets in 15-minute intervals.Resident astronomers used it through 2006 to plan and update nightly schedules; calibrations were often obtained at morning twilight.Source: §2 · p. 6
- After each night, data were made available to each PI with RA comments, enabling prompt feedback on data quality and future observing strategy.PIs accessed their own program data, while all PIs could access calibration files.Source: §2 · p. 7
3. Early Operations for HET Queue and Observing Support
Early HET operations showed that queue scheduling could interleave science, engineering, and commissioning, but actual operations differed from the original staffing plan. The team revised priorities and support processes to address conflicts and improve service to PIs.
3. Early Operations for HET Queue and Observing Support
- Queue scheduling let engineering and instrument commissioning be rescheduled around science observations, so individual astronomers did not bear the full cost of lost telescope availability.Engineers could release the telescope early or pause work briefly for a high-priority observation when feasible.Source: §3 · p. 7
- HET became a 100% queue-scheduled telescope because PIs did not request dedicated time, requiring a third full-time resident astronomer beyond the unfeasible initial staffing plan.The facility manager had to devote full time to daytime support rather than the planned half-time RA role.Source: §3 · p. 7
- The queue policy shared data and coordinated instrument setups for conflicting target-of-opportunity requests, while non-ToO requests were observed and distributed separately.Since 1999, conflicts involved two gamma-ray bursts, four supernovae, and two extra-solar planets.Source: §3 · p. 7
- Early technical unpredictability led HET to standardize overhead at 10 minutes per requested visit until full science operations made actual overheads more predictable.Some visits initially took twice as long as average because of telescope problems.Source: §3 · p. 8
- PIs’ demand for a web-based program-status interface led to a page showing attempted observations, target status, and used TAC allocation.The underlying reports, Phase II data, and TAC allocations had to be combined from separate sources.Source: §3 · p. 8
- Personalized RA communication about data delivery and quality was overwhelmingly approved by TACs and PIs, prompting additions to Phase II files for standards, calibrations, grouped observations, and synoptic timing.The customer-service approach required RAs to record changes in PI materials and communicate them to other resident astronomers.Source: §3 · p. 9
4. Full Operations with the HET Queue and Observing Support
Full operations combined a modified-priority scheduling aid with resident-astronomer judgment and ongoing program monitoring. Completion varied by priority, while staffing demands and unequal access to observing conditions remained operational concerns.
4. Full Operations with the HET Queue and Observing Support
- In 2006, shutter-open science time was 50.3% of clear science time, while bad weather cost 31.9% of nights and scheduled engineering used 3.2%.The remaining 49.7% of clear science time included overheads, mirror alignment, calibrations, and problems.Source: §4 · p. 10
- TACs found that blending priority classes within a program let critical targets be observed while giving PIs flexibility; successful P4 programs used many widely distributed targets with exposures under 600 s.HET typically received two to five new proposals per trimester from time held back for unexpected science such as targets of opportunity.Source: §4 · p. 10
- The expectation of equal partner access across observing conditions was not met: two smaller partners had under 1% of submitted targets in bright time in 2006, with no formal policy to address the inequity.By trimester’s end, bright time was largely dominated by P4 targets from the two larger partners.Source: §4 · p. 10
- PIs rejected 3.1% of science-collecting operating hours in 2006, and there was one disagreement over reobserving at no cost from 1999 to 2006.Borderline observations made up 5.3% of charged observations in 2006.Source: §4 · p. 10
- Staff attrition could leave operations inefficient because rehiring and training took about six months, motivating backup RA coverage and a fourth telescope operator.Science operations used eight FTEs, 40% of total HET FTEs and 27% of its budget.Source: §4 · p. 11
- A feasibility tool helped PIs judge whether proposals were completable, but it did not reveal conflicts with other programs, especially overcrowded sky regions.Queue conflicts were assessed by comparing expected visits across sidereal-time bins.Source: §4 · p. 11
- Queue monitoring combined program completion estimates, sidereal-time target histograms, and RA choices based on availability, priority, synoptic constraints, and current conditions.The modified-priority algorithm suggested a nightly plan and helped predict target order, but RAs still made decisions as conditions changed.Source: §4 · pp. 12–13
- Lower-priority targets had better completion under the modified-priority method than under subjective RA-specific nightly decisions, while highest-priority completion depended on weather and program feasibility.The completion analysis covered median program rates from 2004–2006; the modified-priority system began in 2005 and was further adjusted in 2006.Source: §4 · p. 13
5. Review and Status: Summary of what does and does not works well
The HET queue has been especially successful for target-of-opportunity, synoptic, and wide-sky surveys, and has accommodated engineering work alongside science. Its effectiveness depends on active PI and TAC engagement, staffing continuity, and managing limits on instrument configurations.
5. Review and Status: Summary of what does and does not works well
- HET science return was strongest for target-of-opportunity, synoptic, and wide-sky surveys; most incomplete programs were pencil-beam surveys, while most failures stemmed from technical problems, not queue methods.Queue flexibility also let staff start or pause engineering work around high-priority science, making early operations more productive.Source: §5 · p. 14
- Customer-oriented operations involve PIs in planning and follow-up: they can contact RAs, access nightly data, revise Phase II information, and request repeats of compromised observations.These steps address many queue-scheduling concerns only when PIs examine data and provide timely feedback to RAs.Source: §5 · p. 14
- Phase II and mid-trimester survey response rates were 22% and 28%; 56% of responses were positive without criticism, while Web documentation and Phase II each received 19% constructive comments.Instrumentation received 6% constructive comments; the authors considered feedback moderately useful but the low response rate insufficient to repeat surveys every trimester.Source: §5 · p. 15
- 35.8% of accepted CCD shutter-open time in 2006 went to poor-conditions P4 targets, which also filled gaps between high-priority observations.The policy charged half the nominal shutter-open time and no overhead, incentivizing PIs and TACs to submit poor-weather targets.Source: §5 · p. 15
- Queue operations need a replacement plan for critical staff, ranging from interim coverage to an extra telescope operator in the rotation.The authors describe the extra operator and resident astronomer expense as small but crucial.Source: §5 · p. 15
- PIs need working knowledge of HET capabilities, but the missing living repository of facility and instrument performance information leaves them relying on peer communication to judge feasibility.The authors suggest one additional full-time employee to document capabilities or divide that work among resident astronomers.Source: §5 · pp. 15–16
- The authors recommend knowledgeable TAC members and early observatory-staff participation in allocation decisions, since some non-partner-approved programs were poorly suited to HET and unlikely to finish.Monthly TAC reports and retained allocation for unforeseen opportunities also supported institutional productivity.Source: §5 · p. 16
- Lower-priority LRS programs using less-subscribed configurations had considerably lower completion rates, despite roughly equal median rates for high-priority programs.Only two of four LRS grisms could be mounted per night; the authors warn that specialized upgrades can reduce queue flexibility.Source: §5 · pp. 16–17
6. Next Generation of Queue Scheduling
The next-generation work focuses on automating manual queue tasks and improving software interfaces for PIs, TACs, and resident astronomers. The planned integrated database and role-specific tools aim to reduce workload and support customized access to operational and program information.
6. Next Generation of Queue Scheduling
- The planned software upgrades target manual queue work, including checking program completion, editing queued targets, creating programs and allocations, and generating calibrations.The current process for checking allocations requires manually placing remaining targets on hold.Source: §6 · p. 17
- Medium-term planning has been more sensitive to variable weather than to PI program changes, and the authors expect greater PI control not to alter that balance.Current PI change requests are uncommon outside synoptic programs.Source: §6 · p. 18
- The upgrades require replacing separate night-report and tab-delimited data systems with an integrated database.The selected upgrade was a MySQL database with HTML, PHP, JavaScript, and AJAX interfaces.Source: §6 · p. 18
- The prototype provides flexible, graduated access to program and operational data according to program status or allocation-committee membership.Customized web applications are envisioned for planning, administering, evaluating, and interacting with queue programs.Source: §6 · p. 18