1 NAREGI Grid Middleware and the Operational Issues Kento Aida, Kazushige Saga National Institute of Informatics K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
2 Outline � introduction � NAREGI grid middleware � overview � lessons learned from beta version (through NAREGI-EGEE interoperaiton) � standards in the middleware � deployment � Cyber Science Infrastructure (CSI) � deployment plans � Grid Operation K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
3 NAREGI Project Overview � R&D project for grid middleware � funded by MEXT (Ministry of Education, Culture, Sports, Science and Technology) FY2003 – FY2007 � collaboration of national labs, universities and industry in the R&D activities (IT and Nano-science Apps.) � the Grid layer in the Cyber Science Infrastructure (CSI) K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
4 Project Goals � middleware development � R&D in Grid Middleware and Upper Layer � a prototype of future Grid Infrastructure for scientific research in Japan � testbed � grid computing testbed for nano-science application � international collaboration � OGF � EGEE K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
5 Organization Ministry of Education, Culture, Sports, Science and industry （ MEXT) Operation Collaboration And Collaboration Center for Grid Research and Development Cyber Science (National Institute of Informatics) Supercomputing Technology Industrial Association for Promotion of Infrastructure （ CSI) Deployment Coordination and Project Leader: Dr. K.Miura Operation Committee Computing and Communication Centers Grid Technology (7 National Universities) Research Center etc. Grid Middleware (AIST), Joint R&D Grid Middleware JAEA Integration and And Upper Layer R&D Operation Group TiTech, Kyushu-U, SuperSINET Unitization Osaka-U, Kyushu- Tech., Collaboration Fujitsu, Hitachi, Computational Nano Center NEC ITBL (Institute for Molecular science) Dir.: Dr. F.Hirata Joint Research Collaboration R&D on Grand Challenge Joint Research Problems for Grid Applications （ ISSP, Tohoku-U, AIST, Inst. Chem. Research, KEK etc.) K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
6 NAREGI Software Stack Grid- -Enabled Nano Enabled Nano- -Applications Applications Grid Grid Visualization Grid Visualization Grid PSE Grid PSE Grid Grid Programming Programming Libraries Grid Workflow Tool Grid Workflow Tool Libraries Data Grid Data Grid - GridRPC - GridRPC - GridMPI GridMPI - Super Scheduler Super Scheduler Information Service Information Service （ NAREGI implementation + Globus 4) WSRF （ WSRF NAREGI implementation + Globus 4) Grid VM Grid VM High- -Performance & Secure Grid Networking, Certification Performance & Secure Grid Networking, Certification High SuperSINET Research NII IMS etc Organizations Computing Resources K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
7 Job Execution Scenario Application requirement definition CA/RA Input files Work- MPI flow … Co-allocation source RISM FMO SMP machine PC cluster IMPI 64 CPUs 128 CPUs Job GridMPI b: Deployment a: Sign-on c: Edit Output files Portal PSE WFT 8: Visualization GVS User 1: Submission cert. VOMS Information 2: Resource discovery Super Service MyProxy Proxy cert. Scheduler (User DN, VO) 3: Reservation (Co-Allocation) 9: Accounting GridVM GridVM GridVM 5: IMPI starts 6: MPI job starts 4: Reservation Local Local Local Scheduler Scheduler Scheduler Network monitor 7: MPI init. 2: Monitoring IMPI RISM FMO Job Server Job GridMPI DataGrid Site ρ Site α Site μ Grid File System K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
8 Programming � GridRPC Numerical IDL Library FILE � RPC on the grid IDL Compiler Client 4. 3. invoke generate connect � API standardization by Executable back 2. interface Remote Executable GRAM 1. fork OGF reply Interface interface MDS Information retrieve request � Ninf-G LDIF File Client Server side side � a reference implementation of GridRPC API Cluster A: Cluster B: � GridMPI YAMPII IMPI YAMPII � MPI library on the grid � MPI communication IMPI server between parallel systems on the grid K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
9 User-Level Grid Tools & PSE � Grid PSE � support for compilation and deployment � execution support � Grid Workflow � workflow language � GUI K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
10 Workflow based Grid FMO Simulations of Proteins fragment monomer density dimer total energy data calculation exchange calculation calculation njs_png2005 njs_png2005 njs_png2006 njs_png2006 njs_png2007 njs_png2007 njs_png2008 njs_png2008 njs_png2009 visuali- njs_png2009 zation njs_png2002 njs_png2010 njs_png2002 njs_png2010 input data njs_png2003 njs_png2011 njs_png2003 njs_png2011 njs_png2004 njs_png2012 njs_png2002 njs_png2004 njs_png2012 njs_png2002 NII Resources njs_png2057 njs_png2057 Data component dpcd054 dpcd054 dpcd055 dpcd055 dpcd056 dpcd056 dpcd052 dpcd052 IMS Resources dpcd053 dpcd056 dpcd057 dpcd053 dpcd056 dpcd057 source: Prof. Aoyagi (Kyushu Univ.) K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
11 Adaptation of Nano-science Applications to Grid Environment GridMPI NII IMS (Super-Sinet) Electronic Structure FMO RISM FMO RISM in Solutions Solvent Distribution Electronic Structure Analysis Analysis Data Transformation Grid Middleware Grid Middleware between Different Meshes MPICH-G2, Globus FMO RISM Reference Interaction Site Model Fragment Molecular Orbital method source: Prof. Aoyagi (Kyushu Univ.) K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
12 NAREGI Data Grid Environment Grid Workflow Job 1 Job 2 Job n Data Grid Import data Data 1 Data 2 Data n Components into workflow Data Access Place & Job 1 register data Management on the Grid Job 2 Metadata Assign metadata Construction Grid-wide DB Querying to data Meta- Meta- Job n Meta- Data Resource data data data Management Data 1 Data 2 Data n Grid-wide Store data into distributed file nodes File System K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
13 VO Service Decoupling VOs and RO3 Client Resource Providers Client Client VO- APL2 VOMS Research VOMS Org (RO)1 VO- APL1 SS IS RO2 VOs & SS IS Users VO- RO1 VO- RO2 SS IS VOMS VOMS IS SS Resource IS IS Providers GridVM GridVM GridVM GridVM GridVM GridVM IS IS IS IS IS IS Policy Policy Policy Policy Policy Policy • VO-R01 • VO-R01 • VO-R01 • VO-R01 • VO-R02 • VO-R02 • VO-APL1 • VO-APL1 • VO-APL1 • VO-APL2 • VO-APL2 • VO-APL2 α .RO2 β .RO2 ν .RO2 A.RO1 B.RO1 N.RO1 Grid Center@RO1 Grid Center@RO2 K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
14 Standards in NAREGI Grid-Enabled Nano-Applications ACS Grid Grid PSE Grid Visualization Programming Data -Grid RPC Grid -Grid MPI Grid Workflow CIM GridRPC GridFTP JSDL OGSA Distributed -EMS OGSA Information Service GFS MPI 1.0/2.0 Super Scheduler -DAI OGSA- RUS WSRF （ （ NAREGI implementation + Globus 4) NAREGI implementation + Globus 4) WSRF XKMS X.509 Grid VM VOMS High-Performance & Secure Grid Networking : web service modules WSRF K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
15 Roadmap of NAREGI Grid Middleware FY2008 FY2003 FY2004 FY2005 FY2006 FY2007 ＯＧＳＡ /WSRF-based R&D Framework ＵＮＩＣＯＲＥ -based R&D Framework Utilization of NAREGI-Wide Area Testbed Utilization of NAREGI NII-IMS Testbed Apply Component Prototyping β Ver. Version Technologies to NAREGI Middleware 1 ． 0 Release Nano Apps and Components Evaluation Release Evaluation of α Ver. Deployment of β Version In NII-IMS Development and Testbed Integration of Evaluation of β Version α Ver. Middleware By IMS and other Development and Collaborating Institutes Integration of β Ver. Middleware α Ver. Evaluation on (Internal) NAREGI Wide-area Development of Testbed Verification OGSA-based Middleware & Evaluation Of Ver. 1 K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
16 Lessons Learned from NAREGI Beta � interoperability � NAREGI developed EGEE-NAREGI island as an activity of GIN � Bilateral information exchange � Bilateral job submission � Bilateral file exchange � Interoperable security properties � toward production use � coexistence of multi-type jobs � reservation/non-reservation/local batch jobs � bulk jobs � ease of installation � stability… K e nto Aida, Natio nal I nstitute o f I nfo rmatic s
GIN-info: NAREGI � EGEE EGEE NAREGI 1. Create proxy file Information 2. Aggregate celldomain info Service BDII 3. Create SQL View https://xxx/xxx 4. Query NAREGI info. /w the proxy e.g. GlueCEInfoTotalCPUs = NRG_ClusterJobQueue. NumberOfCPUsPerNode * NumberOfNodesPerQueue 5. Write down the results to plain text file 6. Convert the results to Glue Schema ldif file 7. Store the Glue info to BDII
Recommend
More recommend
