LIGO and GriPhyN

1
LIGO and GriPhyN
Caltech
USC/ISI

• Carl Kesselman
• Ewa Deelman
• Sachin Chouksey

Roy Williams Peter Shawhan Albert Lazzarini Kent
Blackburn
2
LIGO Virtual Data Grid
LIGO Hanford
MIT
LIGO Livingston
Caltech
3
Pictures
4
LIGO Overview

• To detect gravitational waves predicted by
  Einsteins theory of relativity.
• Waves created from
• Rapidly spinning asymmetric compact objects eg.
  pulsars
• Mergers of black holes, neutron stars -- gamma
  bursts?
• Supernova collapse (and post-collapse R-modes)
• Starquakes in neutron stars
• Primæval GW from early universe
• Installation in Louisiana and Washington State
• Other projects Virgo (Italy), GEO (Germany),
  Tama (Japan)

5
Physics and Virtual Data
Each source type has different freq/time
charactaristics ? different virtual data
requirements
6
LIGO Data

• Time series, 16 kHz down to 1 Hz
• 1000 channels
• Gravity wave
• seismic, magnetometer, microphone, engineering,
• Power spectra, cross spectra, sonogram,
• Events
• Each has time and SNR (signal to noise ratio)
• Candidate inspiral events from matched filter
• Other pattern matching
• Examine the big ones, histogram the small ones

7
LIGO vs NVO vs HEP

• Processing
• Network
• NVO is federation, LIGO is coincidence
• In both cases registration is important

detector
events
filter reduce
RDBMS
datacube
pattern matching
LIGO, NVO
1D,2D
8
Data Access

• LDAS (Ligo Data Analysis System)
• Script (Tcl) based
• Primarily for onsite analysis and archiving
• Talks to local files, Hanford/Livingston
• Has a filtering language
• stack-based, like Matlab
• functions eg power spectrum, fft, resample
• LARS
• Can talk to local, HPSS files
• Can do subsetting
• Simple program, simple protocol

9
Activities

• Virtual Data
• A Language Architecture
• Naming is Time/Channel/Transformation
• Smart subsetting
• B LigoVista web display
• Archive at-a-glance drill-down
• Replication
• A Realtime mirror
• Hanford/Livingston/Virgo nternational coincidence
  study
• B Fault-tolerant mirror
• from Caltech to UW/Milwaukee
• Pulsar search
• backfill on Condor farms

10
1A Virtual Data Language

• Virtual Data
• Times Channels,
• eg 668371475-668372499 (GPS time, seconds since
  1981)
• L0 PEM-LVEA_SEIS, L1LSC-AS_Q
• Transformations to Virtual Channels
• eg. L0PEM-EX-SEISX- L0PEM-LVEA-SEISX /
  resample(L1LSC-AS_Q, 1, 64)
• Use LARS and Globus Replica Catalog
• For each requested data value, need to
• Determine if it is instantiated if so, where if
  not, how to compute it
• Plan data movements and computations required to
  obtain all results
• Execute this plan

11
1A Virtual Data Architecture
Power spectrum of seismic channels
P. Charlton
12
1B Virtual Data LigoVista

• Overview of archive in a web browser
• Time based
• month, week, day, hour, ....
• Machine state
• lock status, science run, event density, ....
• Drill down to raw data
• Transformed data
• Uses Virtual Data framework
• compute on the fly and cache

13
2A ReplicationRealtime

• Look for coincidence in Ligo-Virgo data

LDAS analysis
event stream
Interferometer, seismic. etc
Interferometer, seismic. etc
streaming bidirectional transatlantic replication
LIGO
Virgo
14
2A Whistlers from Lightning?
15
2B Replication Fault-tolerance

• Master archive replicated at LSC sites
• Fault tolerant

GridFTP
HPSS_at_Caltech
Beowulf_at_UWiscMilwaukee
Replication API
Replica Catalog
Penn State UT Brownsville
16
3. Pulsar Search
Frequency modulation by Earth movement

• Infinite computation on finite data!
• Tiny, precise signals in deep noise
• Search on 4D space sky position, pulsar
  frequency, spindown,

0.5 TByte image
Condor farms
6 months data
UT Brownsville UW Milwaukee Caltech CACR UW
Madison
frequency bins
Frequency (Hz)
Time (months)
frequency
frequency
time
time
17
(No Transcript)