Published on November 29, 2007
Using Condor for Large Scale Data Analysis within the LIGO Scientific Collaboration: Using Condor for Large Scale Data Analysis within the LIGO Scientific Collaboration Duncan Brown California Institute of Technology LSC Condor Usage: LSC Condor Usage The LIGO Scientific Collaboration is an international effort to detect gravitational waves Condor is heavily used by the LSC for data analysis 8 LSC Condor Pools with 2000 CPUs California Institute of Technology, LIGO Hanford Observatory, LIGO Livingston Observatory, UW-Milwaukee, University of Birmingham (UK), University of Cardiff (UK), Massachusetts Institute of Technology, AEI Over 100 users from various LSC institutions Scientific Results Producedwith Condor: Scientific Results Produced with Condor Four LSC working groups analyze data for different sources Pulsar Analysis Group Setting upper limits on the strength of periodic gravitational waves using the first science data from the GEO600 and LIGO detectors (Phys. Rev. D69 (2004) 082004) Limits on gravitational wave emission from selected pulsars using LIGO data (Submitted to Phys. Rev. Lett.) Inspiral Analysis Group Search for binary neutron star coalescences in the Local Group of galaxies (in preparation) Search for Gravitational Waves from Primordial Black Hole Binary Coalescence in the Galactic Halo (in preparation) Burst Analysis Group Upper Limits on Gravitational Wave Bursts in LIGO’s Second Science Run (in preparation) Stochastic Analysis Group Upper Limits on a Stochastic Background of Gravitational Waves (in preparation) How do we use Condor?: How do we use Condor? Using clusters of jobs to perform parameter sweeps Use Condor DAGMan to manage complex workflows LIGO data comes from three detectors Construct DAGs to describe “coincident” and “coherent” workflows Written a set of Python classes to allow simple construction of complex DAGs Working with Pegasus group (ISI) to use Condor-G to access GRID computing resources LSC Software Stack is built on top of VDT Example of a LIGO Inspiral DAG: Example of a LIGO Inspiral DAG Things we like…: Things we like… Standard Universe job checkpointing DAGMan workflow management Robustness for long running jobs and DAGs User priority load balancing Level of support from Condor Team Things we would like…: Things we would like… Reduce the load on our negotiators Many short jobs (~ 1 hour) on large clusters can stress the negotiation cycle Scheduling jobs submitted by DAGMan is intensive Expansion of Standard Universe support Separation of standard universe file I/O from checkpointing Ability to run compiled Matlab code in standard universe Database back end to Condor Quill sounds very interesting!