AIST Grid activity: Brief updates

1
AIST Grid activity Brief updates
Satoshi Sekiguchi Director, Grid Technology
Research Center, AIST, JAPAN s.sekiguchi_at_aist.go.
jp
2
AIST Grid Technology Research Center (GTRC)

• METI(MITI) / AIST Lab, Established Jan. 15, 2002
• Director Satoshi Sekiguchi
• 22 researchers Posdocs becomes 50 later in this
  year
• 130mil 10mil awarded for bldg., equipment,
  RD and operations
• Will become focal point of Grid research in Japan
• Not just RD, but serve as central CA,
  consortiums, ApGrid principal cite, etc.
• Collaboration w/Institutions and other Grid
  projects

3
Grid Architectures and AIST contributions
Social Framework
Education/training/workshop
Drug Design
Bio Infoma
HEP DG
100M JPY
Applications and Portals
Agriculture
110M JPY
ActiveSheet, BioGrid, Drug Design, HEPGrid,
ThaiGrid Portal, CFD, Chemical Eng.,
Environmental and Ecology Preservation,
QC Portal
High Level Grid Middleware
Programming Tools Ninf-G, Omuni-RPC, Nimrod DPL,
MetaCompiler Simulators Bricks and GridSim, Data
farm Resource Brokers Nimrod-G
Grid RPC/Ninf-G Grid Data Farm
180M JPY
1,250M JPY
Low Level Grid Middleware
Security, Process, Storage, QoS, etc Globus,
Jxta, CPM, GrACE, OGSA
1,500M JPY
Grid Fabric
Clusters, MPPs,VP, SMPs
AP partners have Computers PC, WS, Clusters,
Supercomputers SCE manager WAN Internet, APAN,
GrangeNet, APII, Tsukuba WAN Instruments KEK
Accelerators,
1 USD 129 JPY
4
Tsukuba Advanced Computing Center(????????)
National Institute of Advanced Industrial Science
and Technology (?????????)
5
Supercomputer Installation History
1981.4 ? RIPS(Research Information Processing
System/Station) is established in Tsukuba.
FACOM M-200 is installed. Computer network by
optical fiber is placed. 1984.4 ? FACOM M-380
and IBM 3081K are introduced. 1988.3 ?
Supercomputer CRAY X-MP/216 (Theoretical peak
performance470MFLOPS) is introduced.
Ethernet LAN is spread over the
laboratories. 1993.3 ? Supercomputer is
upgraded to CRAY C916. (Theoretical peak
performance 16GFLOPS) 1993.6 ? FDDI network
over laboratories is introduced. 1999.3 ?
Ethernet LAN is upgraded to Gigabit
Ethernet. 1999.3 ? Supercomputer is upgraded to
HITACHI SR8000. (Theoretical peak performance
512GFLOPS) IBM RS/6000 SP
(Theoretical peak performance252.8GFLOPS) is
introduced. 1999.4 ? RIPS is reformed to
TACC(Tsukuba Advanced Computing Center) 2001.4
? TACC is reorganized as a part of National
Institute of Advanced
Industrial Science and T echnology
(HITACHI SR8000)
(CRAY C916)
(CRAY X-MP/216)
(IBM RS/6000 SP)
6
Possible ApGrid Nodes

• PC Clusters
• 64cpu Linux cluster (AIST, Japan)
• 108cpu QC portal cluster (AIST, Japan)
• 64cpu Linux cluster (TITECH, Japan)
• 216cpu Linux cluster (Osaka U., Japan)
• 8cpu Linux cluster (Tsukuba U., Japan)
• 8cpu Linux cluster (Doshisha U., Japan)
• 8cpu Linux cluster (RIKEN, Japan)
• 32cpu Linux cluster (NCHC, Taiwan)
• 20cpu Linux cluster (NTU, Singapore)
• 14cpu Linux cluster (NUS, Singapore)
• 16cpu Linux cluster (KU, Thailand)
• 10cpu Linux cluster (KMIT, Thailand)
• 16cpu Sun UE Cluster (AIST, Japan)
• 256 Alpha Cluster (AIST, Japan)
• And more

• Supercomputers
• NEC SX-6i (RIKEN, Japan)
• Cray SV1 (AIST, Japan)
• SGI Origin 2000/32 CPU (AIST, Japan)
• SR8000 (AIST, Japan)
• SGI Origin 3200 (NUS, Singapore)
• Cray SV1 (NUS, Singapore)
• Access Grid nodes
• TACC Delivery Grid
• 7 sites
• Devices
• MEG at Kansai Center, AIST

7
Testbed Developments (lessens from ApGrid)

• Security Infrastructure
• Based on GSI
• The ApGrid Testbed runs CAs and issues
  certificates for users and resources.
• The ApGrid Testbed allows multiple root CAs.
• Each country/organization/project could run its
  own CA and these could be root CAs on the ApGrid
  Testbed.
• The ApGrid Testbed is also planning to run an
  ApGrid root CA which is signed by a commercial CA
  such as Verisign.
• Information Services
• Based on GRIS/GIIS
• GrADS could be a good example

8
Resource Donation Guideline

• Required Software
• Globus Toolkit should be installed on every
  donated resource.
• Following information should be provided via
  GRIS/GIIS
• Host specific information (hostname, hardware
  architecture, etc.)
• Installed software (name, version, install
  directory, etc.)
• Detailed information could be better support for
  users.
• Announcements
• When the resource becomes ready, send an email to
  the appropriate MLs.
• Create login accounts
• Every user should be appropriately authenticated.
• Every user should provide his/her user
  certificate signed by a trusted CA.
• Maintenance
• Donated resources should be monitored
  appropriately.

9
Other issues

• Administrative work
• DNS domain (apgrid.org, pragma.org ?)
• Mailing list (pragma-discussion_at_npaci.edu, )
• Home Page (www.apgrid.org, pragma.sdsc.edu)
• Technical issues
• Cluster utilization on the Grid
• Most contributed resources are cluster systems.
• Supposed problems
• private IP address handling, MPICH-G2, etc.
• IPv6/IPsec enhancement
• Firewall consideration
• Standardization required -gt Cluster BOF by Phil