Astro-DISC: Astronomy and cosmology applications of distributed super computing

1
Astro-DISCAstronomy and cosmology
applicationsof distributed super computing
2
Purpose
3
Universe

• The visible universe is large
• 28 billion light years
• 10 million galaxy superclusters
• 25 billion galaxy clusters
• 350 billion large galaxies
• 7 trillion dwarf galaxies
• 30 sextillion (3 1022) stars

4
Data sets

• Sky surveys
• Billions of objects(stars, galaxies, )
• Multiple imagesof the same object

5
Data sets
The sizes of modern survey and simulation
datasets are between 1 and 100 billion objects.
OPPORTUNITIES FOR MAJOR IMPACT
Even larger sets are coming soon.
Their analysis requiresdistributed computing.
6
Astronomers vs. computer scientists
CAN WORK TOGETHER
7
DISC Cloud clusterat Carnegie Mellon
Sixty-four nodes

• Each node
• Eight 2.83GHz cores
• Four 1TB disks
• 16 GByte memory

10GBit / second network
8
Specific problems

• Friends of FriendsIdentification ofgalaxy
  clusters

• Correlation functionsAnalyzing distribution of
  distances between galaxies

IMPLEMENTED AND TESTED
9
More problems

• Quasar detectionIdentifying quasars based on
  the five passband fluxes

• Particle historyTracking the history of
  particles in astrophysics simulations

10
Future work

• Distributed computation for other standard
  astronomy problemsDensity distribution,
  photometric calibration, asteroid detection,...

• General-purpose astronomy toolkitMassive
  spatial indices of celestial objects,integrated
  with distributed algorithms