160 likes | 331 Views
JRA3 B. Guillerminet on behalf of the JRA3 project 22 January 2008. Kick-Off Meeting 21-24 January 2008 Goteborg (Sweden). Outline. Participants Objectives of JRA3 Portal Jobs launching Data access Codes orchestration Visualization Service administration
E N D
JRA3 • B. Guillerminet • on behalf of the JRA3 project • 22 January 2008 Kick-Off Meeting 21-24 January 2008 Goteborg (Sweden)
Outline • Participants • Objectives of JRA3 • Portal • Jobs launching • Data access • Codes orchestration • Visualization • Service administration • Connections with the others activities • Plans • Road map • Risks assessment
WHO • Participants: • PNSC: 24 months (N Meyer, M Owsiak, B Palak, M Ptocennik) • UEDIN: 8 months (L Smith, J Hein) • Chalmers: 4 months (P Strand, T Svedberg, D Nilsson) • IPP: 8 months (D Coster) • CEA: 16 months (M Airaj, B Guillerminet) • ULP: 8 months (E Sonnendrucker, R David, P Navarro) • Structure • JRA3 leader & coordinator of JRA3/JRA4: B Guillerminet (CEA)
Objectives • What
Objective: Portal • Requirements: • single access point (based on ROAM, Shibboleth, SAFE …?) • access to applications (based on Migrating Desktop): • Data (codes & simulations) mining tool (catalogues + database) • Simulation parameters editor (Java based) • KEPLER: workflow (Java based) • Post-processing tools (SCILAB, VISIT …) • user queries • Accounting, monitoring • Tasks: • Specify & realize the Web portal • Integrate our applications in the portal • Who? • PNSC (24 months) • UEDIN (8 months) • Chalmers (?) • IPP (?) • CEA & ULP participate to the integration & tests of the applications
Job transfer via FTP Data transfer via FTP Job scheduling via JCL Proxy Objective: Jobs launching • Requirements: • Complex workflows: loops, various directors, sequential & parallel execution • Long duration simulations: jobs could be launched at each time step (batch) and some must be kept alive (call + notification: interactive jobs) • Need a Java API for EGEE (gLite based) & HPC (UNICORE 6): temporary certificate, job submission, get job status, get job output + message oriented middleware (call + notification) • Integration in the workflow engine (Kepler) • Status: • EGEE: Globus version (already in Kepler) • gLite: in progress • Migrating Desktop as a “gateway” between Kepler and GRID/HPC infrastructure • Needs: • Access to EGEE (gLite + VO fusion) & HPC (unicore 6) infrastructures • Support + investigation for the non-batch part (interactive jobs?) • Who? • CEA (6 months / 16) • PNSC
Objective: Data access • Requirements: • Access to external data (EGEE/HPC): open, get, put … a data structure (allocatable array) • Fast access: in-memory data during the simulation (avoid memory to file + file transfer + …) • Integration in the UAL (Many languages implementation: C/C++, F90, Java, …) • Must be defined clearly: connection with SA3 for SAFE • Needs: • Access to EGEE (gLite + VO fusion) & HPC (unicore 6) infrastructures: SA1 & SA2 • Investigation with the biomedical & astronomy projects (EuroVO_AIDA.pdf @ 4th …) • Support: SA3 • Who? • CEA (4 months / 16) • Chalmers (?) • IPP (?)
Objective: Codes orchestration • Requirements: • Codes launching (EGEE/HPC) and coupling from the workflow engine: use the previously provided libraries • Distributed workflow: part of the workflow will run on the gateway, on EGEE and on HPC • Integration of codes: JRA1, JRA2 deliverables • Building a workflow based on ITM & JRA1/2 codes • Needs: • Investigation of various workflow engines: Gridway, … => collaboration or support • Collaboration with UCSD (Kepler team): designing tool for distributed workflows • Who? • CEA (6 months / 16) • IPP : building the workflow?
Objective: Visualization • Tasks: • Integrate the new visualization tools (based on VTK, Visit) in the code platform: • Developments of specific actors • Used in a workflow (during the simulation) • As post-processing tool • Integration of the compression algorithm (deliverable from JRA4)? actor or as a library • Needs: • Collaboration with JRA4: deliverable, training … • Who? • CEA (<1 months / 16) • ULP (?)
Objective: Service administration • Requirements: • Deliverable from SA3: • User administration • Resource management • Accounting • Reporting • Tool based on SAFE (UEDIN) • Task • Integration in the portal • Needs • Training on SAFE and support from SA3 • Who? • PNSC (24 months) • UEDIN (8 months)
Connection with the others activities • NA1, NA2, NA3 • Management: bi-weekly, f2f (3 months), GA (1 year) • Training on GRID, HPC, SAFE, Migrating Desktop • Dissemination • SA1, SA2, SA3 • Login & CPU time to develop & test our software • Help • Implementation of the service administration facilities
Connection with the others activities • JRA1, JRA2 • JRA1, JRA2 provide several codes • Execution using the API • Needs for orchestration • Data access (ITM data structure)? • JRA4 • JRA3 provides Kepler + tools • Development of actors and integration of VTK, Visit, … • Actor for data compression
2008 2009 2010 Plans 2008 • Road map • Work is mainly in 2008 & 2009 2008
Plans 2008 • Deliverables 2008
Plans 2008 • Risks assessment • Jobs launching: • only the batch mode is available => simulations will be less efficient • Data access • No external data => could use file transfer (GridFTP …) => hidden to the user • Orchestration • No way to run a part of Kepler on the GRID/HPC => use Gridway and build a tool to convert the Kepler workflow to Gridway 2008
Work methods • Define precisely the work plan (JRA3 parallel session) • Email as a communication tool • Additional collaborative tools (Savannah, Wiki …)? • Meetings: PSG (2 weeks) + F2F (3 months) => rely on NA1 management • Mobility?