Grid Computing Activities Within the LIGO Scientific Collaboration

1
Grid Computing Activities Within the LIGO
Scientific Collaboration

• Scott Koranda
• UW-Milwaukee
• skoranda_at_uwm.edu
• On behalf of the LIGO Scientific Collaboration

OSG Meeting Sep. 9, 2004 Boston
2
First, some terminology...

• LIGO Laser Interferometer Gravitational-wave
  Observatory
• LIGO

3
First, some terminology...

• LSC LIGO Scientific Collaboration
• LSC

28 (or more) other institutions
4
Second, who is here...

• Albert Lazzarini
• Director of Computing
• LIGO lab and Caltech
• Scott Koranda
• Research Scientist
• UWM
• Erik Katsavounidis
• MIT faculty
• LIGO lab and MIT
• John McNabb
• Research Scientist
• PSU

• Junwei Cao
• Research Scientist
• MIT

5
LSC DataGrid

• 6 US sites 3 EU sites (Birmingham, Cardiff, AEI)

• total of 5.2 Tflops CPU now
• addition 3 Tflops early 2005

6
LSC DataGrid Details

• Grid Services Deployed
• deployed using LSC DataGrid Server
• built on top of VDT 1.1.13
• includes LSC-specific Grid tools
• GSI-OpenSSH
• logins to cluster head nodes
• soon tunnel all CVS access
• GRAM
• jobmanager-fork at all sites
• jobmanager-condor at 8 sites
• jobmanager-pbs at PSU
• GridFTP server
• access to cluster filesystems
• RLS
• cataloging user virtual data
• deployed at UWM and CIT

• Not Deployed or Used
• MDS
• no strong use case yet
• Monitoring tools
• only modest use of Ganglia
• No SRM concepts
• NFS doing the job for now
• no web services or WRSF

7
LSC DataGrid Details

• Authentication Authorization
• Currently issue X.509 certs via DOEGrids CA
• part of the iVDGL VO
• interest in perhaps running LSC CA
• Update grid-mapfiles by hand!
• presently no centralized VO management
  infrastructure
• this already a problem
• expect by 2005 will need to deploy

8
LSC DataGrid Details

• Lightweight Data Replicator
• bulk replication of data to compute sites
• data discovery
• client tools and API
• Matlab library access
• built on top of
• Globus RLS
• Globus GridFTP
• pyGlobus and python used as glue
• deployed at all 9 sites
• to date more than 50 TB replicated to 9 sites
  across 3 continents

9
LSC Partner in GriPhyN

• GriPhyN Grid Physics Network
• GriPhyN research will enable the development of
  Petascale Virtual Data Grids (PVDGs)
• US CMS
• US ATLAS
• SDSS
• LIGO Scientific Collaboration
• Computer Scientists (Globus, Condor)
• LDR developed at UWM as part of GriPhyN
• Virtual Data Toolkit (VDT)
• one stop shopping for DataGrid software

10
LSC Partner in iVDGL

• International Virtual Data Grid Laboratory
• The iVDGL is a global Data Grid that will serve
  forefront experiments in physics and astronomy
• Hardening, testing, support for VDT
• LDR, LSC DataGrid Client/Server packaged with
  Pacman from iVDGL

11
LSC Parnter in Grid3

• Persistent Grid!
• 2736 CPUs (and growing)
• Virtual Organizations (VOs)and applications
• US CMS
• US ATLAS
• SDSS
• LSC
• BTeV
• Other applications
• SnB Bio-molecular analysis
• GADU/Gnare Genome analysis

12
LSC DataGrid Plans

• LIGO now in full production mode and we have data
• S1, S2 analyzed with S3 in the can
• 6 month run is being scheduled for 2005
• any and all Grid effort must be evolutionary and
  support current analysis efforts with minimal
  perturbations
• we simply dont have the FTE effort to do pursue
  any other strategy...
• Focus on applications
• GW pulsar search workflow is being reworked
• Binary inspiral search workflow is being
  generalized
• Both leverage the Virtual Data System (VDS) from
  GriPhyN
• Goal is to saturate all available LSC resources,
  then move out to wider Grid
• Grid-enabling DMT

13
LSC DataGrid Plans

• Complete/extend the use of virtual data catalogs
  for efficient (re)utilization of data
• Publish data as they become available
  automatically
• Develop LSC specific publishing API
• Continue LDR development
• New metadata scheme
• Administrator friendly bells and whistles
• Enhance/extend persistent LSC DataGrid for the
  collaboration
• VO management tools
• Add sites (Tier 3 at UT-Brownsville, Salish
  Kootenai College)

14
The LSC DataGrid and OSG

• The LSC will continue to watch with interest the
  development of the OSG
• Questions
• We have O(0) effort to contribute...how can we
  contribute?
• We cannot evolve quickly in certain directions
• SRM is a good example
• no effort to port application, no effort to
  deploy SRM
• How can our applications leverage OSG resources?