TrendsTrends inin SoftwareSoftware forfor largelarge astronomyastronomy projectsprojects G.Chiozzi, A.Wallander – ESO, Germany K.Gillies – Gemini Observatory, La Serena, Chile B.Goodrich, S.Wampler - National Solar Observatory, Tucson, AZ J.Johnson, K.McCann – W.M.Keck Observatory, Kamuela, HI G.Schumacher – National Optical Astronomy Observatories, La Serena, Chile D.Silva – AURA/Thirty Meter Telescope, Pasadena, CA
1 AspectsAspects analyzedanalyzed z TimelineTimeline z ChallengesChallenges z ArchitectureArchitecture z FrameworksFrameworks z DevelopmentDevelopment methodologiesmethodologies z TechnologicalTechnological implementationimplementation z HWHW platformsplatforms z OperatingOperating systemssystems z ProgrammingProgramming languageslanguages z User Interfaces. User Interfaces. 2 Keck TimelineTimeline
VLT/VLTI 1990
Gemini N/S 1995
2000 LSST 2005 ALMA
2010 ATST
2015
TMT 2020 E-ELT 3 ChallengesChallenges ofof newnew projectsprojects
z SynchronizedSynchronized multiplemultiple distributeddistributed controlcontrol loopsloops (wave(wave frontfront control)control) z MultiMulti--levellevel offoff--loadingloading schemesschemes z FaultFault detection,detection, isolationisolation andand recoveryrecovery (E(E--ELTELT M1:M1: 10001000 segmentssegments withwith actuatorsactuators andand sensors)sensors) z OperationalOperational efficiencyefficiency (TMT(TMT requirement:requirement: onon targettarget inin <5<5 minutes).minutes). 4 ArchitectureArchitecture
z All major facilities in operation: three-tier architecture z High-level coordination systems z Low-level real time control computers (LCUs) z Devices with limited degree of intelligence z Fairly independent sub-systems: slow correction offloading z Wave front control (adaptive optics and interferometry) introduces new requirements: z Distributed real time synchronization and feedback z Significant physical separation z Systems of systems, often heterogeneous z LCUs role is eroded on both sides.
5 FrameworksFrameworks
z A uniform software framework has Frameworks adopted: a value in simplifying development • Keck and Gemini: EPICS, RTC and maintenance • ESO Paranal and La Silla: z Isolate application from VLT CCS middleware providing a layer of • ALMA and other projects: ACS common services •ATST: ATSTCS z Separation between technical and functional architecture now formally adopted. Common services: • Connection z Component based architectures emerged as particularly useful in •Event distributed systems • Command z Sharing the technical framework • Logging would allow sharing functional • Persistent store components . components . • Error handling 6 DevelopmentDevelopment methodologiesmethodologies andand modelingmodeling techniquestechniques z Our constraints: z Multi-year observatory design periods z RevReviewiew structure and process imposed by funding agencies is oriented to a waterfall approach z Floating requirements z Methodology evolution: z Mid ’80s/ mid ‘90s: Structured programming z Mid ’90s/ beginning 2000: Object Oriented and UML (pragmatic approach) z Now: SysML, agile methodologies: z Requirement management and traceability z Integration in a coherent system model as seen from different disciplines .
7 HardwareHardware platformsplatforms z In most existing observatories: z High level coordination → general purpose WS z Real time → Local Control Units (often VME) z DevDevicesices attached directly to VMEs z Many more options are available now: z High level coordination → Personal Computers z (Soft) Real time → PC with real time OS z Intelligent devices on ETH or industrial buses (CAN) z (Hard) Real time → DSPs and FPGAs z Clusters for raw computing power z Virtualization under evaluation. Trend for the future? .
8 OperatingOperating systemssystems
z The 1990s z Proprietary UNIX z Proprietary RTOS (VxWorks dominating) z The turn of the century: open source z Linux z Real Time Linux z And now? z Questioning Linux z Solaris re-emerging z Open source to stay (Solaris) • OS neutrality z MsWindows (and OPC)? • Real time Java • QNX z Other players? • LabVIEW and LabView-RT •PLCs • FPGAs and DSPs . 9 ProgrammingProgramming LanguagesLanguages
Language Keck VLT z The core language(s): C 251050 246738 z Mid ’80s/ mid ‘90s: C domination C++ 0 84400 z Mid ’90s/ beginning 2000: C++ Capfast 130116 0 takeover Tcl/Tk 9408 81657 z Now: Java explosion, C++ Others 118144 64136 decline, C holds Total 508718 476931 z The glue: from Tcl/Tk to Python and over z LabVIEW’s role growing z We have to cope with: z Different languages for different purposes z Highly distributed systems .
10 UserUser InterfaceInterface z A challenging area. Growing complexity. z We are comfortable with Engineering UI development z We do not have skills for good Operator UIs z Java and Tcl/Tk the most used. z GUI builders are not adequate z Rapid prototyping: necessary, but with a dark side z We cannot afford specialized UI development teams .
11 ConclusionConclusion
New facilities are NOT scaled up versions of existing ones. Paradigm changes may be required z Analysis of control system evolution in observatories is on-going z We have identified clear common trends z We aim at: z Sharing lessons learned z Identifying areas for cooperation z Sharing architectural elements and infrastructure z Cooperation is made easier by international collaborations and the open source movement .
12 Questions?Questions?
The authors represent just a subset of the projects in astronomy. Many more colleagues in the astronomical observatory community have given their ideas and time as we have developed this paper.
Web Links ESO www.eso.org – Email: [email protected] W.M.Keck http://www.keckobservatory.org Observatory Gemini http://www.gemini.edu Observatory ALMA http://www.alma.cl ATST http://atst.nso.edu
LSST http://www.lsst.org Thirty Meter http://www.tmt.org Telescope 13