LSST Project and Survey: Some lessons learned Tony Tyson University of California Davis LSST Chief Scientist 1
Unprecedented control of systematic errors − 30 trillion photometric measurements leads to unusually small statistical error − Each LSST field of view contains over 10 million objects; each 30 second visit contains 500,000 objects with S/N > 10. − Statistical errors average down like root N; systematics don’t − LSST cosmology and time-domain science drivers require unprecedented precision for astrometry and PSF shape − This leads to unprecedented control of systematic errors 2
3
LSST Project Build facility, conduct 10-year survey, process, archive, and serve images and data products Peta-scale Data Management + 8.4m Telescope Science and EPO user Interfaces 3.2Gpix Camera 4
The 189 sensor LSST focal plane constructed from 21 rafts Raft Electronics Board (REB) with Custom Integrated circuits (ASPIC and CABAC) Corner rafts with guiders and WFS not shown Raft Electronics Cage 4K x 4 K Science Sensor Raft Sensor Assembly 5
Lesson learned #1 • CCD R&D began in 2002 • In 2017 we are hard at work characterizing the CCDs so that data management can correct for “instrument signature” systematics • Lesson: Detectors for precision astronomy take a long time to develop. 6
Data Management must transport, process, archive and serve 15 Tb of raw data / night – 500Pb in 10 years Algorithm Design Petascale Computing Design Gigascale Network Design Petascale Database Design 7
Petascale Computing, Gbps Networks The computing cluster at the LSST Archive at NCSA will run the processing pipelines. • Single-user, single-application data center • Commodity computing clusters. • Distributed file system for scaling and hierarchical storage • Local-attached, shared-nothing storage when high bandwidth needed Archive Site and U.S. Data Access Center NCSA, Champaign, IL Long Haul Networks to transport data from Chile to the U.S. Base Site and Chilean • 2x100 Gbps from Summit to La Serena (new fiber) Data Access Center • 2x40 Gbps for La Serena to Champaign, IL (path La Serena, Chile diverse, existing fiber) 8
Database and Science User Interface Massively parallel, distributed, fault-tolerant relational database . • To be built on existing, robust, well- understood, technologies (MySQL and xrootd) • Commodity hardware, open source • Advanced prototype in existence (qserv) Science User Interface to enable the access to and analysis of LSST data • Web and machine interfaces to LSST databases • Visualization and analysis capabilities 9
LSST From the User’s Perspective − A stream of ~1-10 million time-domain events per night, detected and Level 1 transmitted to event distribution networks within 60 seconds of observation. − A catalog of orbits for ~6 million bodies in the Solar System. − A catalog of ~37 billion objects (20B galaxies, 17B stars) , ~7 trillion Level 2 observations (“sources”), and ~30 trillion measurements (“forced sources”), produced annually, accessible through online databases. − Deep co-added images. − Services and computing resources at the Data Access Centers to Level 3 enable user-specified custom processing and analysis. − Software and APIs enabling development of analysis codes. 10
11
12
13
14
Lesson learned #2 • Coordinated management of all parts of the project is critically important. • Lesson: Communication is key. Frequent “all hands” and “joint technical” meetings. 15
Lesson learned #3 • LSST Data Management construction (only): 417 person-years + $55 million hardware, 10 years. • Euclid Data Management construction (only): 750 person-years + 50 million euro, 10 years • Gaia DPAC construction (only): 1400 person-years • Lesson: Data Management is a BIG effort and must begin early in a project. 16
LSST Operations Plan and Total Survey Cost Operations: 140 FTE’s, $40m/ yr (2014USD) $10m/year from EPO Community International Partners Software 4% 8% Project Management Data Management 15% 52% Observatory Operations 21% Total LSST Survey cost: $1.3 billion
Euclid CSS-OS WFIRST LSST
Recommend
More recommend
