Transcript Computational Astrophysics at the Kavli Institute for Particle

Computational Astrophysics
at the
Kavli Institute
for
Particle Astrophysics and Cosmology
at
Stanford
Roger Blandford
9 v 2012
Kavli IPMU
1
High Performance Computing @ KIPAC
• Truism that steadily increasing computational
power has transformed science in general and
astrophysics in particular
• High performance computing contributes to:
–
–
–
–
–
–
Simulation of complex physics under current paradigm
Optimization of telescope design
Exploration of model space
Data management, analysis, archiving and mining
Explanation of discoveries
Public dissemination of results
• Recent example of each type of computing
9 v 2012
Kavli IPMU
2
Simulation of Complex Physics
under Current Paradigm
• Dark matter clumping in expanding universe
• Crucial for understanding:
– Missing dwarfs problem
– Direct detection of WIMPs
– Indirect detection of g-rays
• Abel, Hahn, Kaehler have implemented a
new approach to dark matter simulations
following trajectories in 6D phase space
• Testing and comparison with 3D results
9 v 2012
Kavli IPMU
3
Warm Dark Matter Simulation
9 v 2012
Kavli IPMU
4
Optimizing Telescope Design
• Telescopes are typically designed for both
specific goals and discoveries
• e.g. LSST (2014 start?; 2020 operate?)
– Dark energy through weak lensing
– Light from distant star
•
•
•
•
•
Deflected by intervening gravitational field
Distorted by atmosphere
Reflected by moving mirrors, refracted by thick lenses
Detected and counted by noisy CCD
Analyzed using new algorithms
=>w(a)
– Peterson, Chang, Bard… are building simulator
9 v 2012
Kavli IPMU
5
LSST Simulation
9 v 2012
Kavli IPMU
6
Exploration of Model Space
• Complex physical processes have to be
modeled phenomenologically to tease out
empirical rules
– e.g. how do we associate luminous galaxies with dark
matter and gas distribution
• Busha,Wechsler, Kaehler adapt Bolshoi
simulation and compare with Sloan survey
– Visually indistinguishable
– Compare measurable correlation functions
• Understand rules in terms of basic physics
9 v 2012
Kavli IPMU
7
Bolshoi-SDSS Comparison
9 v 2012
Kavli IPMU
8
Data Management, Analysis,
Archiving and Mining
• Telescopes produce data challenges
• e.g. Dubois manages Fermi data pipeline
–
–
–
–
Event processing in 15 min
Alerts, triggers
1600 CPUs, 4PB disk, tapes
Back up on campus; 1200 CPU system in Lyon
• LSST
– 20 TB per night=>60 PB raw data, 15 PB for catalog
– =>300PB data volume; >150 Tflops
9 v 2012
Kavli IPMU
9
Fermi MISSION ELEMENTS
•
•
GPS
msec
Large Area Telescope
& GBM
DELTA
7920H
-
• Telemetry 1 kbps
•
Fermi Spacecraft
TDRSS SN
S & Ku
•
•
S
-
•
GN
•
Schedules
Mission Operations
Center (MOC)
GRB
Coordinates Network
LAT Instrument
Science Operations
Center (SLAC)
Fermi Science
Support Center
Schedules
Alerts
White Sands
HEASARC
GBM Instrument
Operations Center
Data, Command Loads
9 v 2012
Kavli IPMU
10
Explanation of Discoveries
• Unexpected is expected in astronomy
• Many astrophysical phenomena have no
credible (or many incredible) explanations
• e.g. X-ray quasi-periodic oscillations in
stellar black hole systems ~ 300 Hz, 3:2?
• McKinney, Tchekhovskoy, RB simulated
accretion onto black hole with strong field
–
–
–
–
3D RMHD, >106m, geometries initial conditions
Efficient, quasi-stable jets, extract spin energy
Outflows, winds, Jet-Disk Oscillation
Relativistic radiative transfer underway
9 v 2012
Kavli IPMU
11
9 v 2012
Kavli IPMU
12
Public Dissemination of Results
• Education and Public Outreach is important
part of KIPAC mission
• Staff, postdocs and students regularly
present shows, lead tours, visit schools…
• Pierre Schwob Computing and Information
Center hosts 3D theater and Hyperwall
• Analysis AND outreach
• New graphics, rendering tools, hardware
– GPUs, suitcase system
•
9 v 2012
Kavli IPMU
13
Third Grade in 3D
9 v 2012
Kavli IPMU
14
Summary
• Truism that steadily increasing computational
power has transformed science in general and
astrophysics in particular
• High performance computing contributes to:
–
–
–
–
–
–
Simulation of complex physics under current paradigm
Optimization of telescope design
Exploration of model space
Data management, analysis, archiving and mining
Explanation of discoveries
Public dissemination of results
• Increasingly, these functions are combined
in strongly coupled activities
9 v 2012
Kavli IPMU
15
Congratulations
9 v 2012
Kavli IPMU
16
Reionization (Alvarez et al)
9 v 2012
Kavli IPMU
17
Dark matter streams (Hahn et al)
9 v 2012
Kavli IPMU
18
Large scale structure (Abel et al)
9 v 2012
Kavli IPMU
19
Clusters (Wu et al)
9 v 2012
Kavli IPMU
20
Hyperwall (Adesanya…)
9 v 2012
Kavli IPMU
21