1. Open Grid Forum 27 Banff, Canada Requirements for Future Cyber-Infrastructure Workshop Tuesday, October 13th, 2009 LIGO Cyber InfrastructureKent BlackburnLSC / LIGO / Caltech LIGO-G0900945
2. Brief description of the project. Where are the project participants from? How is the e-infrastructure provided? What level of interoperability is required: data, compute, service? Workshop Questions:
3. Scientific Mission: Directly detect the oscillations of space-time predicted by Einstein’s General Theory of Relativity Two Observatory Sites: Hanford, WA & Livingston, LA L-shaped arms 4 kilometers long on a side. Largest high vacuum system on the Earth’s surface. Able to measure length changes on one one-thousandth the size of a proton over the 4 km distance Instrumentation records roughly one and a quarter Terabytes per day Funded by the National Science Foundation LIGO: Laser Interferometer Gravitational-Wave Observatory
5. Partners in a Global Network LLO LHO VIRGO GEO600 TAMA300 AIGO
6. The LIGO Scientific Collaboration (LSC) is a dynamic group of approximately 700 scientists from approximately 70 institutions worldwide who have joined together in the search for gravitational waves from the the most violent events in the universe. The LIGO Data Grid (LDG) is a collection of computing resources owned and managed by the LSC Provides computing infrastructure necessary for analysis of LIGO data in the search for gravitational waves. Several sites are (will be) part of the Open Science Grid. LIGO Scientific Collaboration
7. LIGO Data Grid The LIGO Scientific Collaboration operates 9 clusters, with a total of approximately 17,000 cores in total. Condor is deployed on all LDG sites. AEI Hannover Cardiff AEI Berlin Syracuse LIGO Hanford UW Milwaukee LIGO MIT LIGO Caltech LIGO Livingston
8. Architecturally, the LIGO Data Grid relies on a fundamental design choice: Replication & Cataloging of Science Data to all centers as a service Users are not entangled by issues of getting data to their jobs/workflows Satellite tools are needed to provide integrity and interoperability Source Code Version Control and Repositories Data Analysis Tool Distributions compatible with user platforms and skills Web servers, Wikis, Document Databases, Data Analysis Databases, Audio/Video Conferencing, … Cyber-Infrastructure Tools
9. The LIGO Data Grid is scaled to provide full coverage of the Compact Binary Coalescence (CBC) analysis This doesn’t have head room for large new analysis models As a baseline scope, this also supports Burst, Stochastic and a limited (targeted) form of the Periodic analyses LIGO also incorporates additional cyber-infrastructures to assure flexibility and new thinking BOINC Applications Periodic searches could utilize every computer on Earth! Open Science Grid Many common services & software with the LIGO Data Grid More on Cyber-Infrastructures
12. Open Science Grid ranked #2 contributorthrough adaption of BOING to submission mechanism onto grids.
13. Open Science Grid LIGO is a key stakeholder in the OSG OSG is used opportunistically by LIGO to gain additional computing resources for two of LIGO’s analyses: LIGO Applications Opens Science Grid Compact Binary Coalescence Workflows (60,000 DAG nodes) Einstein@Home on the OSG
14. CBC Workflow DAG mini-Cartoon LIGO’s Compact Binary Coalescence Analysis relies on in-house tools to generate workflows with complex interdependencies between jobs and data. A typical analysis run involves tens to hundreds of thousands of DAG nodes requiring roughly a week to run each workflow.
15. The enormous variety of interfaces, services, tools making up the cyber-infrastructure used by the LIGO Scientific Collaboration introduces a challenging authentication/authorization problem Dozens of credentials are need by each participant to be able to conduct daily science mission oriented tasks LIGO is actively pursuing a single Auth(2) solution under the purview of the “Auth-Project” Still under active development Facing many challenges from existing infrastructure policies, requiring adoption of redundant or home grown technologies to support user community Challenges Looking Forward
16. What are we looking for? Rules that are easy to master. Ability to distribute objects/apps transparently around the cyber-infrastructure. Evolution when objects warrant. Sufficiently sophisticated to support complex goals. Have some fun on the way!