Iniciativas GRID en la Red Académica Española GT RedIRIS 2002 Jesús Marco, CSIC
Iniciativas GRID Física de Altas Energías: Desafío del próximo acelerador LHC EU-DataGrid (IFAE, testbed) CCLHC-ES LCG (CERN, participación española) LCG-ES DataTag CrossGrid Aplicaciones Interactivas Testbed Empresas EoI 6 PM
The Challenge of LHC Computing ATLAS CMS Storage Raw recording rate 0.1 1 GBytes/sec Accumulating at 5-8 PetaBytes/year LHCb 10 PetaBytes of disk Processing 200,000 of today s fastest PCs
The Challenge of LHC Computing Researchers spread over all the world! Europe: 267 institutes, 4603 users Elsewhere: 208 institutes, 1632 users
The DataGRID project Project supported by the EU Fifth Framework programme Principal goal: collaborate with and complement other European and US projects Project objectives: Middleware for fabric & Grid management Large scale testbed Production quality demonstrations Three year phased developments & demos Open source and communication Global GRID Forum Industry and Research Forum Main partners: CERN, INFN(I), CNRS(F), PPARC(UK), NIKHEF(NL),ESA-Earth Observation Other sciences: KNMI(NL), Biology, Medicine Industrial participation: CS SI/F, DataMat/I, IBM/UK Associated partners: Czech Republic, Finland, Germany, Hungary, Spain, Sweden (mostly computer scientists) Industry and Research Project Forum with representatives from: Denmark, Greece, Israel, Japan, Norway, Poland, Portugal, Russia, Switzerland Collaboration with US similar GRID initiatives
UI JDL Input Sandbox Replica Catalogue Information Service Job Submit Event Output Sandbox Resource Broker Input Sandbox Logging & Book-keeping keeping Job Submission Service Brokerinfo Storage Element Job Status Output Sandbox Compute Element
Spanish Participation in DataGRID WP6 (TESTBED) 2001: IFAE reports in behalf of the other HEP institutions working in the testbed workpackage of the DataGrid project in Spain (IFCA,CIEMAT,UAM,IFIC) Certification Authority Installation kits Information servers (GIIS) Condor batch system and AFS DataGrid project web sites and mailing-lists for Spain Institution Contact Role Funded manpower IFAE A.Pacheco Testbed site Coordination R.Escribá CIEMAT N.Colino Testbed site CMS grid contact F.J.Calonge IFCA R.Marco Testbed site Top GIIS for Spain Certification Authority O.Ponce IFIC J.Salt Testbed site ATLAS grid contact S.González
The CERN LHC Computing Grid project After the CERN Hoffmann review (2000): Resource implications presented to LHC experiment RRBs in March 2001. CERN Management summary presented to SPC and Committee of Council in March as white paper. Discussions between CERN Management and LHC experiment spokespersons. LHC turn-on schedule agreed machine-experiments CERN/2379 green paper for Council and FC in June Development and Deployment of LHC Computing Grid Infrastructure should be setup and managed as a unified project, similar in some ways to a detector collaboration. CERN is viewed as the institution that should co-ordinate it. There should be a Prototyping Phase in 2002-2004. The scale and complexity of the development is large and must be approached using a Project Structure. Work is needed in the Member State Institutes and at CERN. Human and Material Resources for CERN s part of Phase I are not sufficient and should be funded by additional contributions from interested Member States. AGREED! Spanish contribution includes fellowships at CERN
Spain, 2001: Acción Especial for Local Infraestructure Objective: Initial seed for LHC Computing at each site: Trained Personnel Startup Hardware Trigger participation in: CERN LHC GRID Computing project (IT&collaborations) Collaboration software GRID projects
CG-ES -year project oordinated y Manuel elfino (PIC) EAD = Analysis Fa rm EDS = SW Dev Platform R S G = S W repository GSW = SW Gridification MCF = MC Fabric GVM = VirtualM C farm ETD = Data Transf orm PIC = Gridified Dat a Store SE G = Security Arc hitect CTS = Tech MC Support C D C = Data Chal. Coord. Deliverables to fulfill the EAD, GS W, CDC USC objectives EAD, MCF, GSW, CDC UAM IFCA EAD, MCF, EDS, RS G, SEG, CTS EAD, GV M Stay away from glitz. CIEMAT EAD, ETD, PIC, EDS, GS W IFAE UB EAD,EDS, CT S IFIC EAD, MCF, CTS, CDC Concentrate on deploym ent,mc & analysis Use local Univ.for TT to other disciplines 600 KC HF m aterials contribution to LC G-CE R N
The CROSSGRID project European Project (Cross Action CPA9,6th IST call,v PM) [5 M ] Objectives: Extending GRID across Europe: Testbed (WP4) Interactive Applications (WP1) in: Health Care (vascular surgery) Environment (air pollution, meteorology, flooding...) HEP (Interactive Data Analysis) Partners: Poland (CO, M.Turala), Germany (FZK), Holland, Portugal, Greece...(13 countries, 21 institutions) Industry: Datamat (I), Algosystems (Gr) Spain: CSIC (IFCA, IFIC, RedIRIS), UAB, USC/CESGA Participating in : applications (environment,hep), performance and monitoring, resource management, testbed (CSIC WP leader) Started 1 st March 2002 Q1 Deliverables released! (including all SRS, testbed planning)
CrossGrid WP1 Task 1.3 Distributed Data Analysis in HEP Coordinated by C.Martinez (CSIC) Subtask 1.3.2: Data-mining techniques on GRID ANN(Artificial Neural Networks) main tool for Data-mining in HEP Example of Physics Analysis using ANN
HEP Interactive Application User CAS service Interactive Session Resource Broker 2 3a Replica Manager Portal Authorization DATASET 1 4 5 3b XML Input XML Output 9 6 7 Interactive Interactive Session Interactive Session Interactive Worker Session Interactive Worker Session Worker Session Worker Worker Interactive Session Manager DISTRIBUTED PROCESSING 8 DB Installation Interactive Session Database server
Storage Element as WebService? David Rodriguez, CSIC Current SE in EDG: GridFTP server WebService approach: Passive SE : GridFTP, or /grid, etc... Active SE : SQL QUERY (ResultSet in XML)= SELECT FROM (Three tier: servlet running, like Spitfire) ready! (IBM IDS) ROOT query (does this make sense? Paw query does make sense, implemented...) PROCESSING QUERY (= Agent) : Stored Procedure or XML description (SOAP like?) SQL QUERY ok for NN in HEP PROCESSING QUERY (Agent-like approach) needed likely for SOM
HEP Interactive Portal V.O.Authentication DATASET Resources Monitoring DATASET Dictionary (Classes): Basic Object Derived Procedures Graphic Output/(Input?) Analysis Scripts Alphanumeric Output Work Persistency
Distributed (via MPI) NN training scaling Distributed NN Performance 20000 Total Time in seconds 18000 16000 14000 12000 10000 8000 6000 Serie1 Potencial (Serie1) 644577 events, 16 variables 16-10-10-1 architecture 1000 epochs for training 4000 2000 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 # of computing nodes First checks with nodes at Santander & RedIRIS (Oscar Ponce & Antonio Fuentes): remote configuration: modelling including latency <100 ms needed!
S O M Application for DataMining Adaptive Co mpetitive Learning Downscaling Weather Forecasts Sub-grid details scape fro m nu merical models!!!!!
AtmosphericPattern Recognition Prototypes for a trained SOM. Close unitsin the lattice are associated with similar atmospheric patterns. T 1000m b T 500m b Z, U, V 500m b
CrossGrid Architecture (OGSA in mind) Applications 1.1 BioMed 1.2 Flooding 1.3 Interactive Distributed Data Access 1.3 Data Mining on Grid (NN) 1.4 Meteo Pollution Supporting Tools 2.2 MPI Verification 2.3 Metrics and Benchmarks 2.4 Performance Analysis 3.1 Portal & Migrating Desktop Applications Development Support 1.1 Grid Visualisation Kernel MPICH-G 1.1, 1.2 HLA and others Grid Common Services 3.2 Scheduling Agents DataGrid Job Manager 1.1, 1.3 1.2 3.1 User Interaction Distributed Roaming Services Data Collection Access DataGrid Globus 3.4 Replica Replica Optimization of Grid Manager Manager Data Access 3.3 Grid Monitoring GRAM Replica Catalog GSI Globus-IO GIS / MDS GridFTP Local Resources Resource Manager (SE) Secondary Storage Resource Manager (CE) CPU 3.4 Resource Manager 3.4 Optimization of Local Data Access Tertiary Storage 1.1, 1.2 Resource Manager Scientific Instruments (Medical Scaners, Satelites, Radars) 1.1 Resource Manager VR systems (Caves, immerse desks) 1.1 Resource Manager Visualization tools
CrossGrid WP4 - International Testbed Organisation Objectives Testing and validation for Applications Programming environment New services & tools Emphasis on collaboration with DATAGRID + extension to DATATAG Extension of GRID across Europe
CROSSGRID testbed TCD Dublin PSNC Poznan USC Santiago CSIC IFCA Santander UvA Amsterdam FZK Karlsruhe ICM & IPJ Warsaw CYFRONET Cracow II SAS Bratislava LIP Lisbon CSIC RedIris Madrid UAB Barcelona CSIC IFIC Valencia Auth Thessaloniki DEMO Athens UCY Nikosia
CrossGrid WP4 - International Testbed Organisation Tasks in WP4 4.0 Coordination and management IPJ (Warsaw) K.Nawrocki UvA (Amsterdam) D.van Albada (task leader: J.Marco, CSIC, Santander) FZK (Karlsruhe) M.Hardt Coordination with WP1,2,3 IISAS (Bratislava) J.Astalos Collaborative tools (web+videoconf+repository) PSNC(Poznan) P.Wolniewicz Integration Team UCY (Cyprus) G.Tsouloupas 4.1 Testbed setup & incremental evolution (task leader:r.marco, CSIC, Santander) Define installation Deploy testbed releases Certificates Security Working Group A.Fuentes RedIRIS Testbed site responsibles: CYFRONET (Krakow) A.Ozieblo ICM(Warsaw) W.Wislicki TCD (Dublin) B.Coghlan CSIC (Santander/Valencia) J.Sanchez UAB (Barcelona) E.Heymann USC/CESGA (Santiago) C.Fernandez Demo (Athenas) Y.Cotronis AuTh (Thessaloniki) C.Kanellopoulos LIP (Lisbon) J.Martins
CrossGrid WP4 - International Testbed Organisation Tasks in WP4 4.2 Integration with DATAGRID (task leader: M.Kunze, FZK) Coordination of testbed setup Exchange knowledge Participate in WP meetings 4.3 Infrastructure Support (task leader: J.Salt, CSIC, Valencia) Fabric management HelpDesk Provide Installation Kit Network support: QoS (working group, I.Lopez CESGA) 4.4 Verification & quality control (task leader: J.Gomes, LIP) Feedback Improve stability of the testbed JOINING DataGrid testbed 1.2 in July 2002
ands on GRID @ IFCA (http://grid.ifca.unican.es/)
IFCA Research Institute : University of Cantabria Consejo Superior de Investigaciones Científicas Three main research lines: Astrophysics (XMM, Planck...) Statistical Physics (Lasers, fractals & chaos...) High Energy Physics: DELPHI, LEP (Physics Analysis) CDF, Fermilab (TOF detector & Physics Analysis) CMS, LHC (Alignement & Geant4 Sim, OSCAR) Common Interest: Computing needs: Data Management Advanced Analysis Techniques Optimize resources for infraestructure & manpower
HEP Computing at IFCA Previous experience: DELPHI Fast simulation RPC software for DELPHI on-line Analysis software for DELPHI (NN, IES...) Initiatives: Databases (use of O/R DBMS in HEP) FEDER project with DB software company (Semicrol) GRID Initiatives: DataGRID: testbed site & CA for Spain CROSSGRID: WP1 (HEP appl, meteo), WP2, WP4 (testbeds) Technology transfer with companies (Mundivia, CIC) Participation in testbed of DataTag (CDF) Computing for LHC (CMS)
GRID team in Santander Research line at IFCA ( Univ.Cantabria + CSIC ) staff + contracts + fellowships Expertise: Databases use Testbed issues (cluster installation, security, CA, etc) Applications: Astrophysics Complex systems HEP Meteo Collaboration and support (via projects) on NN, methods: Dpto Matematicas Clusters & MPI: Grupo de Arquitectura de Computadores Network: Centro de Calculo U.C. Companies: Mundivia CIC-SL Semicrol
Resources New IFCA building with support for e-science activities (2002/2003) New Infrastructure: Cluster ~100 IBM servers (100% available for GRID) (dual 1.26 GHz, 640Mb-4GB RAM, 80 GB/server) + 4-way processor gatekeeper Gigabit local backbone Improved network connection: 155 (?) Mbps Santander-RedIRIS (Geant node)
72 Computing Elements. Worker Nodes. 8 Storage Elements IBM xseries 220 1 CPU 1.26 GHz 128Mb+512Mb SDRAM Hard Disk: SCSI 30Gb IDE 60Gb Network: 100 Mbps CDROM, floppy NEXT UPDATES 8 Network Cards 1000Mbps (for Storage Elements,...) Join 1.26GHz CPUs in dual setup Buy new >=1.4GHz CPUs Two machines with 4Gb SDRAM for tests
Remote Automatic Installation Nodes configured for PXE Boot Installation Server: DHCP,NFS, TFTP 1 server for LCFG 1 server for PXE-Linux + Kickstart Help sources: PXE-Linux (from SYSLINUX, http://www.syslinux.org) http://syslinux.zytor.com/pxe.php HOWTO Install RedHat Linux via PXE and Kickstart http://www.stanford.edu/~alfw/pxe-kickstart/
A new IST Grid project space (Kyriakos Baxevanidis) Applications - Links with European National efforts - Links with US projects (GriPhyN, PPDG, ivdgl, ) GRIA EGSO CROSSGRID GRIP EUROGRID GRIDLAB DATAGRID Middleware & Tools DAMIEN DATATAG Underlying Infrastructures Industry / business Science
EoI 6 PM (7 Junio 2002) Proyecto Integrado EGEE (coordinado CERN) CSIC: RedIRIS IFCA (Santander) IFIC (Valencia) IMEDEA (Palma) CAB (Madrid) CNB (Madrid) CBM (?) (Madrid) IAA (Granada) Centros: CIEMAT (Madrid) IFAE (Barcelona) PIC (Barcelona) CESGA (Santiago) IAC (Tenerife) Universidades: U. Cantabria U. Valencia U. Murcia U.A.Barcelona U.A.Madrid U.Complutense Madrid PYMES: CIC-S.L. (Cantabria) GridSystems (Palma)
EoI 6 PM (7 Junio 2002) Red de Excelencia RTGRID (Real Time GRIDs) España: CSIC Univ.Cantabria CESGA CIC-SL Polonia Cyfronet Grecia Univ. Athenas Univ. Thessaloniki Slovakia IISAS Bratislava Cyprus Univ. Cyprus Otras propuestas: CEPBA UPV?...
In perspective GRIDs will help with: Organizational and large scale issues Metacomputing Web Services are commercial OGSA could be the way if performance is ok Interactive Grid will be hard without QoS on networks Several GRID projects with Spanish participation progressing well Need for organization in Spain: Thematic Network + Teams to organize work e-science centers to get local support, administrative organization, dissemination and exploitation (we need companies involved)