Grid and Supercomputing Activities at the ICT Division of CIEMAT R. Mayo (On behalf of the authors) ICT Division – CIEMAT (Spain) IberGrid 2009 Valencia, May 20 th -22 nd 2009
Valencia, IberGrid 2009, May 20 th -22 nd Outline Introduction Grid and Latin America – EELA – EELA-2 Grid and Fusion – EUFORIA – EGEE-III Grid support: EPIKH Grid Domestic Initiatives Supercomputing Activities – Precedents – Euler
Valencia, IberGrid 2009, May 20 th -22 nd Introduction CIEMAT – Spanish Public Organism for Research and Technological Development since 1951 – Structured in several Departments and Divisions Technology ICT Division – Computation activities from the 50s of the previous Century
Valencia, IberGrid 2009, May 20 th -22 nd EELA EELA (EC SSA – 2006/2007) – To bring e-Infrastructures of Latin American countries to the level of those of Europe – Project Coordinator – WP1 (Management) Administrative tools and support – WP2 (e-Infrastructure) ROC – WP3 (Applications) Biomedicine, HEP, e-Learning – WP4 (Dissemination) Tutorials and events – Qualified as ‘Good to excellent ‘ by EC
Valencia, IberGrid 2009, May 20 th -22 nd EELA-2 EELA-2 (EC I 3 – 2008/2010) – To ensure long-term sustainability – Coordinated by CETA-CIEMAT – SA1 (e-Infrastructure) Services provided: WMS, CE, SE, UI, BDII… 130 cores and 6.6 TB – JRA1 (Research) VM – NA3 (Applications) Phylogenetics Fusion Radiotherapy/Medical detectors
Valencia, IberGrid 2009, May 20 th -22 nd EELA-2: VM Virtual Machines – As workload units – Application and VO compliant execution – To face a highly heterogeneous environment – Goals Dynamic VO specific services and software packages Creation of customized virtual clusters for specific apps Overlay a virtual cluster on more than one physical cluster
Valencia, IberGrid 2009, May 20 th -22 nd EELA-2: DKEsG Drift Kinetic Equation solver – LNF – Neoclassical transport for 3D Fusion devices (TJ-II) – DKEs monoenergetic diffusion coefficients per parameter Radius of the magnetic surface, Electric Field Temperature, Density Ions or electrons – Division in minimal tasks that can run on cluster and Grid – DKEsG coefficients of the Onsager transport matrix
Valencia, IberGrid 2009, May 20 th -22 nd EELA-2: PhyloGrid PhyloGrid – Fundación IDEA – Determination of the evolution history of different species Influence between hosts and virus – MrBayes – Workflow in Taverna Web service – Examples Duffy domain HIV HPV
Valencia, IberGrid 2009, May 20 th -22 nd EELA-2: ALICE A Large Ion Collider Experiment – CERN, USC, CEADEN – nucleus-nucleus interactions at LHC energies MC events of a particular quarkonia state Coalescence model of production of D in pp and AA interactions – ALICE VO_BOX – ALiEn SW – Dedicated SE (4.4 TB) – Production (since August 2006) ~ 750 CPU KSI2K hours More than 70 Kjobs
Valencia, IberGrid 2009, May 20 th -22 nd EUFORIA EUFORIA (EC I 3 – 2008/2010) – To enhance the modelling capabilities for ITER sized plasmas through a set of critical applications HPC Grid – Coordinated by Chalmers Tekniska Högskola
Valencia, IberGrid 2009, May 20 th -22 nd EUFORIA EUFORIA (EC I 3 – 2008/2010) – GEM Code Max-Planck-Institut für Plasmaphysik (Garching) and LNF Univ. Edinburgh (HPC version) MPI -f particle code that includes the full dynamics of gyrokinetic ions and drift kinetic electrons Workflow ETS
Valencia, IberGrid 2009, May 20 th -22 nd EGEE-III EGEE-III (EC I 3 – 2008/2010) – Largest multi-disciplinary grid infrastructure in the world – Coordinated by CERN – CIEMAT participates in several Activities
Valencia, IberGrid 2009, May 20 th -22 nd EGEE-III EGEE-III (EC I 3 – 2008/2010) – FAFNER2 code LNF NBI heating of plasmas (TJ-II configuration) MC calculations Able to be executed on Cluster (MPI) Grid (DRMAA and MPI) Workflow ISDEP ETS
Valencia, IberGrid 2009, May 20 th -22 nd EPIKH EPIKH (EC MCA – 2009/2012) – To consolidate e-Science by a well structured programme Latin America Asia Africa – Coordinated by Consorzio Cometa – Grid Schools System Administrators (1 week) Applications (2 weeks) Final workshop (1-2 days)
Valencia, IberGrid 2009, May 20 th -22 nd Spanish Initiatives Spanish e-Science Network (MICINN CAC) – To coordinate and foster the scientific activities in e-Science – Coordinated by UPV – In Grid Apps and e-Infrastructure areas – NGI gLite Grid site (40 cores & 2.2 TB) General, Operations and Training VOs supported – JRU
Valencia, IberGrid 2009, May 20 th -22 nd Spanish Initiatives Spanish e-Science Network (MICINN CAC) – Accredited Group of Support for apps Physics, Fusion, Bioinformatics – I3A-Univ. Zaragoza mitochondrial DNA Phylip & MrBayes
Valencia, IberGrid 2009, May 20 th -22 nd Spanish Initiatives GRIDImadrid – In the framework of Biogridnet – Coordinated by UCM – Devoted to research on middleware techniques – Globus site Ibercivis – BOINC – Dissemination activities (web)
Valencia, IberGrid 2009, May 20 th -22 nd SC: Precedents Computation Activities at CIEMAT – since 1959 – Many scientific fields Fusion, HEP, Environmental Sciences… – Many platforms Vectorial, Cray, Altix, cc-NUMA, X86... – Open to any researcher (not only to the CIEMAT staff) Some examples – Uniac Solid State UCT – 1959 – Univac machines (1110/81 by terminal) – 70s – IBM 3090/150 (LAN) – 1985 – Cray Computers (T3E parallel arch) – 90s – SGI machines (Linux) – XXI Century
Valencia, IberGrid 2009, May 20 th -22 nd SC: Nowadays Some of the previous, still operative A step beyond: clusters – Better value for money – The most extended architecture worldwide – Around 200 processors HPC Grid – Many SW libraries/tools provided
Valencia, IberGrid 2009, May 20 th -22 nd SC: Now and then Euler – Constellation cluster 144 processors Dual Xeon quad-core 3.0 GHz (2 GB per core) RAM memory of 2.3 TB for the 1152 cores 140 GB of storage per node (21 TB as a whole) Interconnected by Infiniband Fully devoted to the execution of jobs X86 architecture – Lustre File System – multi-tier RAID 6+0 disk cabinet (60 TB) – Linpack Test on Intel Chip 5450 R peak = Tflops ; R max = Tflops – To be upgraded in 2010 (70%)
Valencia, IberGrid 2009, May 20 th -22 nd SC: Now and then CIEMAT aims to join RES – Providing Euler resources – Scientific support ICT Research – CIEMAT Data Centre is heterogeneous – Friendly interface for the submission of jobs Maximising the resources – Based on GridWay and Virtual Machines
Valencia, IberGrid 2009, May 20 th -22 nd SC: Now and then
Valencia, IberGrid 2009, May 20 th -22 nd Conclusions CIEMAT has always fostered ICT – Not only internal support, but research too Involved in several Projects Aims to collaborate with other researchers in many disciplines (Grid & HPC)
Valencia, IberGrid 2009, May 20 th -22 nd Alicia Acero, Angelines Alberto, Fernando Blanco, Jorge Blanco, Francisco Castejón, José A. Fábregas, Liubov A. Flores, Manuel Giménez, Pedro González, José Guasp, Raúl Isea, Pedro Ladrón de Guevara, Rafael Mayo, Esther Montes, Antonio Muñoz, Manuel Rodríguez-Pascual, Antonio J. Rubio-Montero, Eulogio Serradilla and Stefano Troiani THANK YOU