1. 6 march 20001 Building the INFN Grid Proposal outline a.ghiselli,l.luminari,m.sgaravatto,c.vistoli INFN Grid meeting, milano

2. 6 march 20002 Contributions from: -Monarc test-bed -Globus tests -Grid workshop and tutorials (Padova CHEP-2000)

3. 6 march 20003 What’s a Grid? GRID is a set of services for -Obtaining information about grid components -Locating and scheduling resources -Communicating -Accessing code and data -Measuring performance -Authenticating users and resources -Ensuring the privacy of communication …. -From “The GRID” edited by I.Foster and C.Kesselman

4. 6 march 20004 HEP computing Key elements: -Data type and size per LHC exp.: RAW(1MB per event, 10 15 B/year/exp.), ESD(100KB per event), AOD(10KB per event) -Data processing: event reconstruction, calibration, data reduction, event selection,analysis, event simulation -Distributed, hierarchical regional centers architecture: Tier 0 (CERN), Tier 1, …desktop

5. 6 march 20005 -Data model: data organized in objects stored in distributed object data base -Desirable ODBMS features: -Object stored and managed with C++ -Client/Server architecture -Data distributed in several data servers -Fault tolerance and data replication and transparent access to the nearest copy…..

6. 6 march 20006 Approaching computing and data Grid Three main working-areas -Identify GRID requirements -Application programming technologies -Grid layout model

7. 6 march 20007 Identify GRID requirements infn-grid WG identified 4 main topics: 1)Wide area Workload management 2)Wide area data management 3)Wide area application monitoring 4)Computing fabric and general utilities for a global managed grid They are included in the present EU-GRID WorkPackages

8. 6 march 20008 -HEP Application programming model Commercial technologies:Java, CORBA… -Need interfaces to GRID services Grid oriented programming: can take full advantages of the grid services User oriented technologies …… Application programming technologies

9. 6 march 20009 GRID Layout Model Data Server CPU Client Layout evolution: Multi-level server hierarchy desktop CERN WAN Condor pool CPU Client

10. 6 march 200010 Modeling the grid Starting from the applications: 1)Traditional appl. running on single computer and with local I/O 2)Client/Server appl. running on multiple computers and with local or remote I/O 3)Grid oriented appl. integrated with grid services

11. 6 march 200011 Grid for traditional applications -Use case: Event simulation -Condor based grid: possible I/O bottleneck, no CPU load balance -Globus based grid: Design and implement High Throughput Scheduler How to handle Appl. I/O: -Modifying the appl., or with DATA mover -Globus/Condor based grid: job submission from Globus client to Condor pool and vice versa. Personal Condor configuration could be an interesting HTB for Globus.

12. 6 march 200012 workplan -Use case : CMS HLT ph.1, ATLAS trigger studies -Deliverable: HTB, DATA Mover -Test bed: Globus pool, Condor pool, PC machines -Schedule: 6-8 months (rough estimate) -FTE: 5 (rough estimate)

13. 6 march 200013 Grid for client/server applications -use case: analysis application with Objectivity 5.2 in multiserver configuration (later on Objectivity can be substituted by Espresso or other…) -Issues to investigate: -How clients and servers can interoperate with the grid -Advance co-reservation and co-scheduling -Monitoring and real-time workload management

14. 6 march 200014 Cnaf PD Mi server Clients 10Mbps Wan link Pisa Genova Clients 10Mbps Wan link Clients 10Mbps Wan link Testbed Layout Garr-b TEN-155 Bologna RM server Other servers or clients 10Mbps Wan link CERN Bari

15. 6 march 200015 Workplan -Use case : analysis applications -Deliverable: -Appl. Programming interface to grid services -Monitoring and real-time workload management -Advance co-reservation and co-scheduling -Test bed: dedicated grid layout with 10Mb as minimum bandwidth -Schedule: 2 years -FTE: 9

16. 6 march 200016 Interactive networked applications -Interactive phisics analysis -Remote monitoring and control -Requirements: High performance rather than high throughput -Workplan: tbd

17. 6 march 200017 Grid oriented applications They are implemented in terms of various application toolkits component, grid services and grid fabric mechanisms -tbd

18. 6 march 200018 Summary schema applicationsBatch traditional appl. Client/Server batch applications Client/Server Interactive applications Grid enabled applications Application oriented Grid Services/ Grid application toolkits HTB (High Th. Broker) Data mover Monitoring tools Real-time workload management HPerformance Scheduler HPerformance data access Remote data access, remote computing, remote instrumentation, Application monitoring Grid Common Services (Resource and appl. independent services) Common Informatio n Base (MDS), Authentication, authorization, policy (GSI) Global Resource Monitoring GRAM (Globus Resource Allocation Manager) Grid Fabric/local Resources Computing resources local schedulers, site accounting ….. Network services Data storageInstrumentation

19. 6 march 200019 Conclusions 1.Grid Fabric: Definition of CPU resources, Network services, Network layout 2.Grid services: Decide the Grid Common Services to adopt. MDS, GSI, GRAM…… 3.Application oriented Grid Services: Identification of the most important Services satisfying the application use cases. 4.Application development using the GRID. 5.Project management organization: Project organization: top – down approach based on application use cases. Coordination in order to identify common services to be shared between the applications. GRID Architecture needs GRID Organization Coordination/integration with the European Grid Project