We will have a NAREGI booth at SC2005 which will be held November
12-18, 2005, in Seattle, WA. For the last two years NAREGI, a Japanese national
Grid R&D project, has been developing a wide range of production-level software
including infrastructure-level middleware, client/user-level tools and applications.
The highlights of NAREGI software include:
- An incarnation of OGSA-EMS's resource and job management
- Reservation based co-allocation of MPI jobs across multiple sites
- Ease of use of NAREGI middleware and client/user tools for developing
large-scale grid applications e.g., nano-science coupled simulation
You are cordially welcome to stop by our NAREGI booth (#1127) on November
14-17, where you will be able to see hand-on demos, talking and asking
questions, if any, to NAREGI technical staff members.
SC/05 NAREGI Demonstration Program(Booth#1127)
1
10:40 - 10:55
13:40 - 13:55
NAREGI Overview Dr. Kenichi Miura (National Institute of Informatics)
2
11:00 - 11:15
14:00 - 14:15
NAREGI Middleware online Demonstration Dr. Satoshi Matsuoka (Tokyo Institute of Technology / NII)
3
11:20 - 11:35
14:20 - 14:35
Grid-enabling of GAMESS-FMO program Dr. Mutsumi Aoyagi (Kyushu Univ.)
4
11:40 - 11:55
14:40 - 14:55
Application to Soft Nanoscience Dr. Akihiro Morita (Institute for Molecular Science)
5
12:00 - 12:15
15:00 - 15:15
Data Integration of Scientific Databases Dr. Hideo Matsuda (Osaka Univ.)
6
Coupled Simulation using NAREGI a Version During intermission of above demonstrations
7
Demonstrations of GridMPI and GridRPC are taking place at the AIST booth(# 723 ).
1) NAREGI Overview The goal of NAREGI is to develop grid middleware for widely-distributed computing resources in advanced research and education. The NAREGI project involves participation by industry, academia, and government institutions. Based on success in building a science grid environment, the project is launching a cyber-science infrastructure to promote R&D in various scientific fields.
2) NAREGI Middleware online demonstration NAREGI resource management middleware is one of the world first reference implementations of GGF/OGSA(TM)-EMS architecture. It supports automatic resource brokering and reservation based co-allocation, which makes it easier for users to run complex jobs that require running on across systems on different sites. In this demonstration, we explain the architecture and how it works with a simple MPI job. (OGSA is trademark of Global Grid Forum)
3) Grid-enabling of GAMESS-FMO program To execute nano applications more efficiently on the grid environment, we decomposed a FMO program into modules to make computational units loosely coupled each other. Each module can use the appropriate computer resource on the grid environment, which the NAREGI middleware selects. We have carried out the calculation on the serum albumin molecule (18,242atoms) on the NAREGI grid environment with Loosely Coupled FMO.
4) Application to Soft Nanoscience A variety of combinations of different computational techniques can be realized in the grid environment. Here two examples are demonstrated in the soft nanoscience:
(1) MD (molecular dynamics)-MO (molecular orbital) combination,
(2) replica MC (Monte Carlo) RISM (reference interaction site model) combination.These techniques are quite powerful to study chemical reaction mechanisms and structures of biopolymers in water.
5) Data Integration of Scientific Databases A large number of databases are available in life sciences, astronomy, etc. Using our Data Grid environment, the data integration of those DBs is performed as: (1) construct metadata for associating all the related data in the DBs, (2) connect the DBs and the metadata, (3) describe dataflow for querying across the DBs, and (4) deploy engines on the environment and execute the dataflow.
6) Coupled Simulation using NAREGI alpha Version This demonstration shows the usage for end users and the internal mechanism of NAREGI grid, by taking a practical and somewhat complex case: large-scale RISM*1-FMO*2 coupled simulation for analysis of electronic structure of protein Lysozyme in water. It is executed as a heterogeneous parallel MPI job on a large number of automatically selected computing resources across multiple sites.
*1 Reference Interaction Site Model *2 Fragment Molecular Orbital Method
