Data Processing & Quality Control Group [ Faculty Meeting Sept. 2010 Data Processing & Quality Control Quality Control Group head: Reinhard.

Data Processing & Quality Control Group [www.eso.org/qc] Faculty Meeting Sept. 2010

Data Processing & Quality Control

• Quality Control Group• head: Reinhard Hanuschik• 7+1 QC scientists (astronomers)• sitting in TOB (with DMO)

Reinhard Hanuschik


• pipeline-process all VLT+VLTI data• certify calibration products, store in archive• process science data, store in archive• create data packages for PIs

• QC mission: ensure that science data can be calibrated, to a known and documented level


• supported:– all Paranal instruments:

11 VLT, 2 VLTI instruments, plus survey telescopes (VISTA, OCAM)

• close interaction with:– Paranal Science Operations – User Support– other groups at DPD (External Data Products, Science Data

Products)– Archive Department


• statistics per month:

– 50.000 processing jobs (20.000 VLT, 30.000 VIRCAM) (+150% in 2 years)

– 6 TB raw data processed– 0.8 TB product data created and archived– 120 SM data packages – 300 DVDs

• processing:– on 13 dual-core blades, plus– on a cluster of 20 dual-core blades (VIRCAM, HAWKI, OCAM)


• how do data get here?

– data transfer Paranal – Antofagasta – Santiago – Garching– microwave link: now at 10 Mbits/sec – sufficient for all VLT(I) data plus VIRCAM calibrations

– most raw files available in archive within minutes

– next step: fibre link Paranal – Antofagasta (EVALSO, 2010/2011)


• how is QC making use of it?

process calibration data: incremental and automatic processing– once per hour: look for new CALIB data – pipeline-process them in the QC way (dynamic associations

within pool of data, fine-tuned processing parameters etc.)– create QC reports– calculate scores– put this information on the web– do this 24 hrs a day, 7 days a week

quasi-real time quality control and feedback– evaluated by:

– QC scientist– Paranal daytime or nighttime astronomer

– QC does review/certification (off-line)


• an example …

life product monitor examples: 2010-07-12 2010-07-13 today

(all fine) (quality problem) ?

http://qcweb.hq.eso.org/GIRAFFE/logs/2010-07-12/status_2010-07-12.html




life product monitor examples: 2010-07-12 2010-07-13 today






rms_fit – rms of fitted dispersion solution


for comparison: healthy situation

HC monitor link: www.eso.org/HC

http://www.eso.org/observing/dfo/quality/GIRAFFE/reports/HEALTH/trend_report_ARC_RMS_HC.html


nlin_acc - number of lines for the dispersion solution


mean_R – mean resolving power


certification / rejection: interaction & analysis by QC scientist

product monitor examples: 2010-07-12 2010-07-13 today



• QC review and certification

– this is done off-line (Mon-Fri, 5/2)– score-based– review all red scores, plus selected green scores– analyze issues– reject or certify– communicate with PSO

• certified master calibrations go into archive


one QC data point = one calibration

product

• trending of all core instrument components

• for complete history of instrument


• quick-look: score version

– concentrates on compliance with predefined thresholds– numerical values replaced by scores: simple and graphical – hierarchical score scheme instrument score


• Health Check monitor:

– central tool for quality feedback to Paranal SciOps– closes the feedback loop: raw – pipeline – scores - HC– up-to-date at timescale of 1 hour– based mostly on CALIB information (daytime, nighttime) – includes also SCIENCE information (image quality;

VIRCAM: zero points)


• The process for SCIENCE data– pipeline-processing: off-line– all pipeline supported instrument modes (40-50)– SM or VM: does not matter– use certified master calibrations– no strict certification process (on-line quality is graded

ABCD by SciOps); feedback about quality problems or uncalibrated OBs given to USD and SciOps

– products are ingested into archive– data packages are stored on the User Portal and available

there for PIs– timescale for User Portal availability to PIs:

– raw data within minutes after acquisition– product data within days (or faster)


the end


• processing level of SCIENCE data– pipelines: originally mainly for quality control and quick look– goal: remove instrumental and atmospheric signatures

• strong push from community towards science-grade quality– sheer volume of raw data– complexity of data reduction

• push pipelines and calibration plans towards science-grade products (SDP)

• to be implemented within the QC scheme: mass production with pipelines, not fine-tuned towards specific scientific goal


the end


• calChecker: automatic interface to check completeness of calibrations– again: feedback scale less than one hour– 24/7– checks completeness of all calibrations formally required

for science, within validity range

calChecker live

Data Processing & Quality Control Group [ Faculty Meeting Sept. 2010 Data Processing & Quality Control Quality Control Group head: Reinhard.

Documents