Enabling Grids for E-sciencE Two Seismology Applications on EGEE D. Weissenbach, G. Moguilny, E. Clévédé, E. Stutzmann and J.-P. Vilotte Institut de Physique du Globe de Paris CNRS - Institut des Grilles (France) www.eu-egee.org Information Society and media NERIES —Edimburgh 1 6-7 November 08 The EU - EGEE Grid project Enabling Grids for E-sciencE Grid projects driven by the CERN LHC collider (∼ 10 PetaBytes/year). Mutalize distributed CPUs and storage: today 80 K CPUs and 15 PB. Beside LHC VOs, others like Earth Sciences (ESR) and Astronomy VOs. Users submit jobs from a User Interface (UI) to a Resource Broker which locates an appropriate Computing Element (CE) upon data proximity (Storage Element or SE) and available CPUs using informations published on BDII systems and VO catalogs. gLite middleware ... NERIES —Edimburgh 2 6-7 November 08 GEOSCOPE noise analysis Enabling Grids for E-sciencE Daily noise averages analysis have been completed for all the GEOSCOPE dataset GEOSCOPE Network (G) VOR KAM ECH SSB WUS SCZ INU TAM KIP UNM HDC FDF MBO HYB ATD BNG MPG TAOE RER PPT SPB PEL COYC ROD NOUC CAN NAM TRIS AIS CRZF PVC DZM PAF DRV real time / near-real time stations temporarily interrupted stations remotely accessible stations planned stations locally accessible stations GMT 2007 Jun 5 13:56:47 • GEOSCOPE archive: 25 years, ≤ 30 stations, ∼ 1TB • ∼ 7600 (short) jobs: perl + FORTAN + (matlab →) octave • processing: over 1 year (GEOSCOPE data–center latencies) NERIES —Edimburgh 3 Next step: compute hourly averages; incorporate as data annotations. 6-7 November 08 Application perspectives Enabling Grids for E-sciencE Grid UI submit ➐ get output ➋ ➎ ➍ execution... ➏ ftp Grid SE netDC • client-server implementation; EG E • interface for direct access to Data Centers from the Grid with minimal latency; ➏ submit E grid Grid Job Grid UI He re? ata tad me ➒ ➎ ➐ req ue st ➑ ➊ ➋ order Data Center NERIES —Edimburgh ➒ Data Service store Grid SE ➍ t ge ➌ ut outp • noise analysis as a proxy for data annotation using higher order extracted informations. 4 Grid Job here ? • improve noise application worflow; ➊ re qu es t (m ail ) • shorter period noise analysis of GEOSCOPE data sets; ➌ EG E E grid 6-7 November 08 Earthquake CMT determination Enabling Grids for E-sciencE From Geoscope From USGS EarthQuake Solomon Islands 2007/4/1 { SEED Data Acquired From Near Real Time Seismic Stations First Estimation of Earthquake Parameters Magnitude Mw = 8.1 Strike=331o Dip=38o Rake=120o M 0 =1.6e28 Nm N E Publish Results Data Pre−processing Results Available Setting Space parameters CMT determination Quality Control Synthetics calculation • 3D space grid (earthquake location) × 1D time grid (source duration) ∼ 1800 points • Process: 600 CPU-hours. • Gridified in 2004 (→ 48h restitution time) since then continuously optimized (→ 8h). Linear inversion Server instantiation Jobs submission Synthetic Seismograms Archive For each grid point, 6 Green functions computed per component for each station. • Extensive job submission on VO CEs; each running job requests grid points until final completion. 5 NERIES —Edimburgh 6-7 November 08 gCMT application perspectives Enabling Grids for E-sciencE From Geoscope From USGS EarthQuake SEED Data Acquired From Near Real Time Seismic Stations First Estimation of Earthquake Parameters Publish Results Data Pre−processing Results Available Setting Space parameters CMT determination Quality Control Synthetics calculation Linear inversion Server instantiation Jobs submission Synthetic Seismograms Archive Speed-up of CMT-solution process Re-engineer and optimize workflow Green function: 80% of CPU time. • definition of a global spatial grid for the Earth upper lithosphere; • improve interface between external data center and the EGEE Grid; • construction of a synthetic Green functions archive; • add data center annotation services for results propagation. NERIES —Edimburgh 6 6-7 November 08 Conclusions and Remarks Enabling Grids for E-sciencE • Grid infrastructures: new capacities for on demand or routine data processing and analysis IF interfaced with the community’s data centers. • Some key issues: – interfaces and services between seismological portals and Grid infrastructures; – Grid must be taken into account in European initiatives for data storage and distribution; – improve latency of data request and retrieval; – add new functionalities for data sets annotation and push back of extracted informations to improve data curation. Problems could be adressed though collaboration between the EKIDNA European proposal and NERIES. 7 NERIES —Edimburgh 6-7 November 08