Google Sky Status and Goals
Data Resources Photo Photo Photo CCD Source FITS FITS FITS JP2 Status 16 bit 8 bit 16 bit 32 bit Range 0.7” 0.5” 1” 0.45” Plate scale 21.0 mag 21.0 mag 21.0 mag 23.5 mag Depth JF  (N) JFN (S) JFN (N) JF  (S) JFN (N) JFN (S) ugriz Bands 40000 40000 40000 8000 Coverage COSMOS PMM DSS SDSS
Data Resources SDSS 8000 sq degrees, CCD, 23 mag, ugriz, 0.5” Pittsburgh (jpegs), FNAL (fits) Scanning across the sky (great circles) Ver1: original 0.25x0.25 degree patches Ver2: individual fields Ver3: original fits images (ugr, gri, riz) Licensing started DPOSS/UKSchmidt All-sky, Photo, 21 mag, JFN, 1” PMM All-Sky, 21 mag, JF, 1”
SDSS DSS COSMOS PMM
Fusion Enterprise Box Delivered (fusion and server) 200 sq degrees ingested and served SDSS reprocessed as compressed JPEG2000 (single resolution) SDSS 280,000 regions, 8000 sq degrees 1300x1900 pixels (11m) DSS 180,000 regions,  2500x2500 pixels (16m) Tested all sky WMAP
Current Status Issues Timing for fusion 350 frames (0.1% of SDSS) 274986 tiles touched 2 days 17 hrs > 2000 days?  2-3 hours to build pyramid of 14 levels Merging assets Limit on number and resolution of assets All assets same resolution, must be broken into individual assets Merged at project level Auto mask -> feathering between assest (dark bands) No mask -> does not build all resolution levels (appear at differnet zooms) Artifacts Detail at different zooms Diffraction spikes etc – what will public accept Image quality at poles
Generic Reduction Pipeline Simplify Astronomer’s Ingestion Assumes World Coordinate System wcsOverlap Calculates overlap between images wcsTanProj Project onto tangent plane  (base image and overlaps) Background subtract Lanczos interpolation (windowed sinc function) imageColor Applies color mapping for input images (Lupton algorithm) I ≡ (r + g + b)/3 R =  r ∗  f(I)/I G =  g ∗ f(I)/I  B =  b ∗ f(I)/I
Rotation and background
Color Definition SDSS DSS JFN DSS JF extrapolate
 
Pipeline Requirements Base Map Processing Estimates (no optimization) Fits reduction pipeline 2-3 hrs per plate per color 10-20,000 hrs Fusion 24-50,000 hours Space Estimates DSS 7 TB (FITS) 1.5 TB (1 arcsec, JP2) SDSS 10 TB (FITS) 1.2 TB (0.45 arcsec, JP2) Optimization Generate initial images on the fusion grid? Optimal input image size? Low priority queues
Actions Defining Coordinate System ICRS (J2000) GIS: -180  to 180 deg  Astro 0 to 360 deg New coordinate grid on sky, new client? Projection North up East left? Color Balancing Image scaling for bright stars  Images appear noisy at low zoom Recreate images at each zoom level Balance between images/surveys Need to optimize overlaps of DSS and SDSS Colors for astrophysics or public?
KML and Catalogs Prototype of press release and catalog query HST, SIAP, DBMS Scaling Querying images/points How many points, images in cache How many kml objects can be served Need for addition astrophysics tags What altitudes to we view a set of points Query based on field of view
Issues Data Sources What catalogs/overlays for public Hubble press releases, ACS Bayer, NGC/UGC, RC3, Clusters, Constellations  Layers Layers of expertise for users
Deploying Google Sky How to get to Beta New Client or New database EC can do this How many base maps Base map transparency? What functionality in initial beta Coordinate system and grid Excluding ads Licensing status What is the process to deploy to data centers
New Features Hierarchy of images Small user – single image (OK) Large user – base map (OK) More than one base map Intermediate user WMS standard interface Fast but needs GS to cache images Persistent and rapid update of images How will things scale
New Features Time/Frequency Domain Real time events Press release (linked via CNN etc) RSS feeds of new events Sharing of events from users (comets, asteroids etc) Imaging surveys Multiple realizations SDSS 25 repeats over 300 sq degree LSST/PanSTARRS 100s over 20,000 sq degrees Equivalent to different wavelengths (Xray, Near-IR) Movie of the dynamic sky Embedding into Google Earth? Provide sky around GE
VO Interfaces SIAP Importing links to kml Catalogs XML tags for astronomy Relation with VOTables? Scaling to 10^6 sources
Short-term Critical Actions Low priority processing queues Need 40,000+ cpu hours  MapReduce (already reduced) Which resources are available Ingestion/Magrathean Enterprise box can store and serve final products (network disk) Enterprise box cant process images

GoogleSky Status at Google

  • 1.
  • 2.
    Data Resources PhotoPhoto Photo CCD Source FITS FITS FITS JP2 Status 16 bit 8 bit 16 bit 32 bit Range 0.7” 0.5” 1” 0.45” Plate scale 21.0 mag 21.0 mag 21.0 mag 23.5 mag Depth JF (N) JFN (S) JFN (N) JF (S) JFN (N) JFN (S) ugriz Bands 40000 40000 40000 8000 Coverage COSMOS PMM DSS SDSS
  • 3.
    Data Resources SDSS8000 sq degrees, CCD, 23 mag, ugriz, 0.5” Pittsburgh (jpegs), FNAL (fits) Scanning across the sky (great circles) Ver1: original 0.25x0.25 degree patches Ver2: individual fields Ver3: original fits images (ugr, gri, riz) Licensing started DPOSS/UKSchmidt All-sky, Photo, 21 mag, JFN, 1” PMM All-Sky, 21 mag, JF, 1”
  • 4.
  • 5.
    Fusion Enterprise BoxDelivered (fusion and server) 200 sq degrees ingested and served SDSS reprocessed as compressed JPEG2000 (single resolution) SDSS 280,000 regions, 8000 sq degrees 1300x1900 pixels (11m) DSS 180,000 regions, 2500x2500 pixels (16m) Tested all sky WMAP
  • 6.
    Current Status IssuesTiming for fusion 350 frames (0.1% of SDSS) 274986 tiles touched 2 days 17 hrs > 2000 days? 2-3 hours to build pyramid of 14 levels Merging assets Limit on number and resolution of assets All assets same resolution, must be broken into individual assets Merged at project level Auto mask -> feathering between assest (dark bands) No mask -> does not build all resolution levels (appear at differnet zooms) Artifacts Detail at different zooms Diffraction spikes etc – what will public accept Image quality at poles
  • 7.
    Generic Reduction PipelineSimplify Astronomer’s Ingestion Assumes World Coordinate System wcsOverlap Calculates overlap between images wcsTanProj Project onto tangent plane (base image and overlaps) Background subtract Lanczos interpolation (windowed sinc function) imageColor Applies color mapping for input images (Lupton algorithm) I ≡ (r + g + b)/3 R = r ∗ f(I)/I G = g ∗ f(I)/I B = b ∗ f(I)/I
  • 8.
  • 9.
    Color Definition SDSSDSS JFN DSS JF extrapolate
  • 10.
  • 11.
    Pipeline Requirements BaseMap Processing Estimates (no optimization) Fits reduction pipeline 2-3 hrs per plate per color 10-20,000 hrs Fusion 24-50,000 hours Space Estimates DSS 7 TB (FITS) 1.5 TB (1 arcsec, JP2) SDSS 10 TB (FITS) 1.2 TB (0.45 arcsec, JP2) Optimization Generate initial images on the fusion grid? Optimal input image size? Low priority queues
  • 12.
    Actions Defining CoordinateSystem ICRS (J2000) GIS: -180 to 180 deg Astro 0 to 360 deg New coordinate grid on sky, new client? Projection North up East left? Color Balancing Image scaling for bright stars Images appear noisy at low zoom Recreate images at each zoom level Balance between images/surveys Need to optimize overlaps of DSS and SDSS Colors for astrophysics or public?
  • 13.
    KML and CatalogsPrototype of press release and catalog query HST, SIAP, DBMS Scaling Querying images/points How many points, images in cache How many kml objects can be served Need for addition astrophysics tags What altitudes to we view a set of points Query based on field of view
  • 14.
    Issues Data SourcesWhat catalogs/overlays for public Hubble press releases, ACS Bayer, NGC/UGC, RC3, Clusters, Constellations Layers Layers of expertise for users
  • 15.
    Deploying Google SkyHow to get to Beta New Client or New database EC can do this How many base maps Base map transparency? What functionality in initial beta Coordinate system and grid Excluding ads Licensing status What is the process to deploy to data centers
  • 16.
    New Features Hierarchyof images Small user – single image (OK) Large user – base map (OK) More than one base map Intermediate user WMS standard interface Fast but needs GS to cache images Persistent and rapid update of images How will things scale
  • 17.
    New Features Time/FrequencyDomain Real time events Press release (linked via CNN etc) RSS feeds of new events Sharing of events from users (comets, asteroids etc) Imaging surveys Multiple realizations SDSS 25 repeats over 300 sq degree LSST/PanSTARRS 100s over 20,000 sq degrees Equivalent to different wavelengths (Xray, Near-IR) Movie of the dynamic sky Embedding into Google Earth? Provide sky around GE
  • 18.
    VO Interfaces SIAPImporting links to kml Catalogs XML tags for astronomy Relation with VOTables? Scaling to 10^6 sources
  • 19.
    Short-term Critical ActionsLow priority processing queues Need 40,000+ cpu hours MapReduce (already reduced) Which resources are available Ingestion/Magrathean Enterprise box can store and serve final products (network disk) Enterprise box cant process images