Документ взят из кэша поисковой машины. Адрес оригинального документа : http://acat02.sinp.msu.ru/presentations/shawhan/shawhan_plenary.pdf Дата изменения: Thu Jun 27 00:45:28 2002 Дата индексирования: Mon Oct 1 20:32:49 2012 Кодировка:

LIGO Data Analysis
Peter Shawhan
(LIGO Laboratory / Caltech)

For the LIGO Scientific Collaboration

VIII International Workshop on Advanced Computing and Analysis Techniques in Physics Research Moscow State University June 26, 2002

ACAT 2002

Peter Shawhan (LIGO/Caltech)


Gravitational Waves
Predicted by Einstein's theory of General Relativity Massive objects, moving at velocities near the speed of light, can distort the geometry of space-time Far from source, appear as transverse quadrupole waves

Time

Dimensionless strain: h = L / L Typical strain amplitude at Earth: < 10-
21

!

Gravitational waves have not been detected ... yet
ACAT 2002 Peter Shawhan (LIGO/Caltech)


Sources of Gravitational Waves
"Inspiral" (orbital decay) of a compact binary system Two neutron stars, two black holes, or one of each One of the most promising sources, since:
Binary neutron-star systems are known to exist The waveform and source strength are fairly well known (until just before merging)


h
"Chirp" waveform
Note: the measured orbital decay rate of the binary pulsar PSR 1913+16 exactly matches the expected rate due to gravitational radiation !
ACAT 2002 Peter Shawhan (LIGO/Caltech)


Sources of Gravitational Waves
Supernova explosion Wave emission depends on asymmetry of explosion "Ringing" oscillations of a newly formed black hole Stellar collapse to form a neutron star Rapidly-spinning neutron star Will radiate if slightly asymmetric Stochastic radiation from the early universe "Unexpected" sources ? This is a new observational science ! Detection of gravitational waves will open a new window on astrophysics and will test the theory of General Relativity
ACAT 2002 Peter Shawhan (LIGO/Caltech)


Gravitational-Wave Detectors
First detectors: resonant aluminum "bars" First built by Joseph W eber in the 1960s Several cryogenic bars are currently in operation and achieve high sensitivity at their resonant frequencies Several large interferometric detectors are now being commissioned: LIGO, VIRGO, TAMA, GEO Use a laser to measure distance Sensitive over a wide frequency range AURIGA detector

The search for gravitational waves is an international effort
ACAT 2002 Peter Shawhan (LIGO/Caltech)


The LIGO Project
LIGO = Laser Interferometer Gravitational-wave Observatory Funded by the U.S. National Science Foundation "LIGO Laboratory" consists of Caltech and M.I.T. The broader "LIGO Scientific Collaboration" includes over 300 scientists from over 30 institutions worldwide Perhaps ~75 are actively involved in data analysis

ACAT 2002

Peter Shawhan (LIGO/Caltech)


LIGO Optical Layout
(Other large interferometric detectors are similar)

Fabry-Perot arm cavity

Basically a Michelson interferometer, with mirrors at the ends of two "arms" Light returning from the two arms interferes at the beam splitter Intensity at photodetector depends on the difference between the arm lengths, e.g. due to a grav. wave Additional semi-transparent mirrors form optical "cavities" which increase power and enhance sensitivity
End mirror

Mode cleaner

Laser Photodetector
ACAT 2002

Servo feedback to mirror positions keeps all cavities in resonance
Peter Shawhan (LIGO/Caltech)


A LIGO Mirror in situ

ACAT 2002

Peter Shawhan (LIGO/Caltech)


LIGO Observatories
Livingston Observatory Louisiana One inteferometer (4 km)

Hanford Observatory W ashington Two interferometers (4 km and 2 km arms)

ACAT 2002

Peter Shawhan (LIGO/Caltech)


LIGO Status
Construction and detector installation is complete All three interferometers have been operated successfully in their complete optical configuration Some servo systems have not yet been commissioned Currently working to reduce noise and improve robustness Have conducted several "engineering runs" To practice collecting and analyzing data First "science run" begins this Saturday ! Will collect data for 17 days Sensitivity is still a few orders of magnitude away from the design goal, but is starting to be of scientific interest

ACAT 2002

Peter Shawhan (LIGO/Caltech)


LIGO Data
Main "gravitational wave channel" (servo signal which measures the arm length difference) is sampled at 16384 Hz Synchronized to GPS time reference Data stream also includes many auxiliary channels Readback channels from the various servo systems Environmental sensors (seismometers, magnetometers, etc.) There is no "trigger" -- channels are sampled continuously and written first to disk, later to tape Total data rate from each interferometer: ~3 MB/sec Gravitational wave channel is only ~2% of the data stream Data to be archived: ~100-200 TB per year

ACAT 2002

Peter Shawhan (LIGO/Caltech)


Gravitational-Wave Data Analysis
Different scientific topics require different analysis methods Searches for (short) transient signals Inspiral and other known waveforms: optimal filtering "Bursts" (transients with unknown waveform): several algorithms Searches for (long) periodic signals Requires integrating over long periods Search for stochastic gravitational-wave background Requires cross-correlating data from different detectors Detector characterization Requires access to auxiliary channels
ACAT 2002 Peter Shawhan (LIGO/Caltech)


LIGO Data Analysis System (LDAS)
Follows a "computing center" model Dedicated hardware Machines are on a "private" network; no access from Internet except to a single gateway machine Main data archive is at Caltech; data is copied to other sites as needed Software environment created specifically for LIGO Remote job submission and result retrieval via gateway Socket-based job submission protocol; no unix login by users Access requires an LDAS username / password LDAS systems at LIGO observatories, Caltech, M.I.T., and a few institutions in the LIGO Scientific Collaboration
ACAT 2002 Peter Shawhan (LIGO/Caltech)


Components of LDAS
Internet Gateway machine Private network Frame Data Interface Data Conditioning MPI Launcher Event Monitor Database Interface
ACAT 2002 Peter Shawhan (LIGO/Caltech)

Data on disk

LDAS Manager Client Web Server

PC

Relational Database


Implementation of LDAS
LDAS components are separate unix processes Run on several different machines (normally) Socket-based job control and data transmission The LDAS Manager controls job scheduling, as well as the sequence of component operations for each job Components are written in Tcl is used for job control, high-level operations CPU-intensive operations in C++, called from the Tc Tcl and C++ interprocess communication, and on data objects are implemented l layer

ACAT 2002

Peter Shawhan (LIGO/Caltech)


Parallel Processing in LDAS
Uses MPI running on a "Beowulf cluster" of PCs LDAS delivers input data to memory on master node User's analysis code is in the form of a shared-object library Must contain a standard set of entry points Loaded at run time, then called Scientific users have control over the parallelization scheme Typically, the data is broadcast to all nodes, which process it differently (e.g. searching for different waveforms) Results are collected on master node, then passed to "Event Monitor" component e.g. candidate event times and waveform parameters

ACAT 2002

Peter Shawhan (LIGO/Caltech)


A "dataPipeline" Job
Standard job type to search for gravitational-wave event candidates

Frame Data Interface Data Conditioning MPI Launcher Event Monitor

Data on disk

LDAS Manager Client Web Server
Instructions Raw data Event candidates
ACAT 2002

PC

Database Interface
Peter Shawhan (LIGO/Caltech)

Relational Database


A "getMetaData" Job
Retrieves table of event candidates (or other info) from the relational database

Frame Data Interface Data Conditioning MPI Launcher Event Monitor

Data on disk

LDAS Manager Client Web Server
Instructions Event candidates

PC

Database Interface
Peter Shawhan (LIGO/Caltech)

Relational Database

ACAT 2002


The "LIGOtools" Software Suite (Client Utilities and More!)
Philosophy Facilitate the sharing of useful software tools Support diverse users in the LIGO Scientific Collaboration Simple installation Can be installed in any directory; doesn't require root privilege Software is distributed as precompiled binaries [for Sun Solaris and Intel Linux] and scripts No software prerequisites and no "build" step Simple update procedure Just type "ligotools_update" and select from the list of available new packages and/or new versions; the script downloads and installs them for you Central web site with documentation, FAQs, links
ACAT 2002 Peter Shawhan (LIGO/Caltech)


LIGOtools Packages
Tcl library providing intuitive interface to execute LDAS jobs* Graphical interface to LDAS relational database* Utilities for remote retrieval of frame data* Graphical user interfaces to view frame data I/O libraries to read frame data and database table data into a C program or into Matlab LIGO Algorithm Library (LAL) ... and more * These utilities can be (and are) "web-patched" to fix bugs, adapt to evolution of the LDAS communication protocol, or add new features without any action on the part of users
ACAT 2002 Peter Shawhan (LIGO/Caltech)


Lessons Learned
Creating a new data analysis system takes a lot of effort! Integrating the parts is a large fraction of the work It's hard to cast an analysis into a different paradigm Analysis code written to run on LDAS's Beowulf cluster cannot trivially be adapted for execution on a grid Right now, with science runs imminent, we can't afford to spend much time pursuing alternative approaches Having multiple computing environments dilutes expertise Number crunching is only one part of the whole analysis Bookkeeping Collecting, summarizing, and interpreting results

ACAT 2002

Peter Shawhan (LIGO/Caltech)


Summary
LIGO is poised to begin collecting scientifically interesting data A customized environment has been developed to support various LIGO data analysis needs We are attempting to address issues of "usability" and support for a widely distributed community of users LIGO data analysis tools and algorithms will continue to evolve

ACAT 2002

Peter Shawhan (LIGO/Caltech)