Present Present and Future and Future of of One
Shared by: rottentees
Present and Future of One of the largest Grid Infrastructure in Europe T. Ferrari*, L. Carota*, L. Gaido**, F. Nebiolo* for the Italian Regional Operations Centre * INFN-CNAF, Bologna **INFN-Torino The Italian Grid infrastructure currently comprises more than forty resource centers, which are hosted and operated by various research institutes and the academy, such as INFN, CNR, ENEA, SPACI and Universities. It is fully integrated with the pan-European Grid infrastructure built and operated under the umbrella of the project EGEE-III, and is currently strengthening its integration with various Italian Grid regional infrastructures. The Italian Grid infrastructure provides a distributed computing platform based on the principle of resource sharing by users from many scientific disciplines, such as Physics, Astrophysics, Biology, Health, Chemistry, Geophysics, Economy, Finance and extensions to other sectors such as Civil Protection and Industry. It has a central role in Grid development and deployment activities in Italy, from sharing its infrastucture and know-how, to establish platforms for further development of Grid applications. Application, Grid Services, Available storage per Virtual The Italian Grid: and Grid Foundation 36 Sites, 39 Virtual Organization, Organization Middleware Relationship 6555 CPUs, 5.2 PB of Storage • 4 PB of disk space Applications • 1 PB of tape storage Higher-Level Grid Services: Workload Management OTHERS Logging and Bookeeping CNR ILC Pisa CNR PROD Pisa ops dteam lhcb cms atlas Data Management INFN ENEA-INFO Frascati ESA-ESRIN Frascati alice virgo argo infngrid biomed INFN-Bari INAF-Trieste babar theophys zeus cdf egrid INFN-Bologna CNR-ITB Bari gridit bio inaf enea compchem INFN-Cagliari SISSA Trieste esr magic ingv plank pamela INFN-Cascina SNS-Pisa libi cyclops compassit euindia euchina INFN-Catania SPACI Cosenza eumed geant4 gear edteam eela Foundation Grid Middleware: INFN-CNAF INFN-Ferrara SPACI Lecce SPACI Napoli na48 auger fusion embrace Security Model INFN-Firenze INFN-Frascati UNI Perugia Computing Element INFN-Genova INFN-Lecce Accounting per VO INFN-LNL Information and monitoring INFN-LNS (last 12 months) INFN-Milano Accounting INFN-Napoli INFN-Padova INFN-Parma INFN-Perugia •The Workload Management System INFN-Pisa INFN-Roma1 (WMS) accepts user jobs, assigns INFN-Roma2 INFN-Roma3 them to a site, records their status and Prog. Operativo INFN-Tier1 INFN-Torino retrieves their output. Nazionale “Ricerca” INFN-Trieste •The Logging and Bookeeping (LB) SCOPE-UNI-Federico II-Napoli PI2S2-COMETA-SCIRE_Napoli PI2S2-COMETA-INGV Roma service tracks jobs managed by the PI2S2-COMETA-Catania PI2S2-COMETA-Palermo WMS. PI2S2-COMETA-Messina CRESCO-ENEA-Casaccia •The Storage Resource Manager CRESCO-ENEA-Frascati CRESCO-ENEA-Bologna provides dynamic space allocation and CRESCO-ENEA-Portici (NA) CRESCO-ENEA-Trisaia (MT) •Italian grid: https://grid-it.cnaf.infn.it/ CRESCO-ENEA Brindisi file management, defining a standard CRESCO-Uni-Salerno •European grid: https://eu-egee.org/ CRESCO-Uni-Tor Vergata interface to allow interoperability CYBERSAR Cosmolab Cagliari •USA grid:https://www.opensciencegrid.org/ between different storage systems. •The LCG File Catalogue maintains mappings between Logical File Names, Computing Jobs supported Storage URL and Grid Unique • The Message Passing Interface (MPI): a commonly dteam 14,30k (29,3%) ops 11,75k (24,1%) IDentifiers. used standard library for parallel programming. lhcb 7,43k (15,2%) atlas 5,78k (11,8%) • The File Transfer Service is a low WMS handles MPI jobs as a set of processes that alice 3,35k ( 6,9%) biomed 3,20k ( 6,6%) level data movement service which run on different Worker Nodes (WN) in the same CE. others 3,04k ( 6,2%) schedules data transfer jobs between • Collections: sets of independent jobs. source and destination sites according • Checkpointable jobs: the WMS saves intermediate to configurable internal policies. job states in the LB; • The Grid Securiy Infrastructure is • Direct Acyclic Graphs: sets of jobs linked by based on public key encryption, X.509 relative dependencies; certificates, and the Secure Soket • Partitionable jobs: sets of independent sub-jobs, Layer protocol. A system called VOMS cms zeus euchina eumed plus an initial pre-job and a final aggregator job; theophys esr infngrid bio (Virtual Organization Management lhcbprd geant4 atlassgm gridit • Parametric jobs: collections where the jobs are egrid compchem libi geant4sgm Services) is used to manage cdf argo identical, but for the value of a running parameter; information about roles and privileges • Interactive jobs: they open a real time connection of users within a Virtual Organization Jobs submitted with a remote host. The job standard streams are (VO). •The Computing Element (CE) is a set redirected from/to the remote host. via resource brokers of services that provide homogenous, managed, and secure access to heterogeneous, remote computing Future Operational Model resources. The future operational model will need to support the •The principal information and management of resources and services of an monitoring service is named GridICE, infrastructure which is expected to quickly grow in it collects information according to scale, in terms of number users, services, sites standard schema and collects operated and middleware stacks supported. The historical data in a persistent storage. guiding lines of this transition will be the autonomy •The accounting system (DGAS) of national Grid, self-functionality and the full perform an accurate Usage Metering integration with the operational model adopted by the related to jobs and users by sensors pan-European Grid infrastructure. installed on Computing Elements.