Noriyuki Kushida

Toward an international sparse linear algebra expert system by interconnecting the ITBL computational Grid with the Grid-TLSE platform

Complex optimization problems are of high interest for Process Systems Engineering. The selection of the relevant technique for the treatment of a given problem has already been studied for batch plant design issues. Classically, most works reported in the dedicated literature yet considered item sizes as continuous variables. In a view of realism, a similar approach is proposed in this paper, with discrete variables for representing equipment capacities, which leads to a combinatorial problem. For this purpose, a Genetic Algorithm was used, which is multiparametric by nature and a grid approach is perfectly relevant to this case study, since the GA code must be run several times, with different values of some input parameters, to guarantee its stochastic nature. This paper is devoted to the presentation of a grid-oriented GA methodology. Some significant results are highlighted and discussed.In the present paper, the methodology of interoperability between ITBL and Grid-TLSE is described. Grid-TLSE is an expert web site to provides user assistance in choosing the right solver for its problems and appropriate values for the control parameters of the selected solve. The time to solution of linear equation solver strongly depends on the type of problem, the selected algorithm, its implementation and the target computer architecture. Grid-TLSE uses the Diet middleware to distribute computing tasks over the Grid. Therefore, extending the variety of computer architecture by Grid middleware interoperability between Diet and ITBL has a beneficial impact to the expert system. To show the feasibility of the methodology, job transfering program as a special service of Diet was developed.

Development of Three-Dimensional Virtual Plant Vibration Simulator on Grid Computing Environment ITBL-IS/AEGIS

Center for computational science and e-systems of Japan Atomic Energy Agency is carrying out R&D in the area of extra large-scale simulation technologies for solving nuclear plant structures in its entirety. Specifically, we focus on establishing a virtual plant vibration simulator on interconnected supercomputers intended for seismic response analysis of a whole nuclear plant. The simulation of the whole plant is a very difficult task because an extremely large dataset must be processed. To overcome this difficulty, we have proposed and implemented a necessary simulation framework and computing platform. The simulation framework based on the computing platform has been applied to a linear elastic analysis of the reactor pressure vessel and cooling systems of a nuclear research facility, the HTTR. The simulation framework opens a possibility of new simulation technologies for building a whole virtual nuclear plant in computers for virtual experiments.Copyright

Element-wise Implementation of Iterative Solvers for FEM Problems on the Cell Processor

A new implementation of the finite element method(FEM) that is suitable for the Cell processor. Since the Cell processors have a far greater performance and lower byte per-flop (B/F) rate than traditional scalar processors, the amount of memory transfer was reduced and a technique was used to hide the number of memory accesses. The amount of memory transfer was reduced by accepting additional floating-point operations without storing the data if the data were required repeatedly. In the present study, such memory access reduction was applied to the conjugate gradient (CG)method. In order to achieve memory access reduction in the CG method, element-wise computation was used in order to avoid global coefficient matrices, which cause frequent memory accesses. Moreover, all data transfer times are incorporated into the calculation time. As a result, the new implementation performed 10 times better than a traditional implementation run on a PPU.

High Speed Eigenvalue Solver on the Cell Cluster System for Controlling Nuclear Fusion Plasma

In this study, we developed a high speed eigenvalue solver that is the necessity of plasma stability analysis system for International Thermo-nuclear Experimental Reactor (ITER) on Cell cluster system. Our stability analysis system is developed in order to prevent damages to the ITER from plasma disruption. MARG2D, which is the main part of the system, analyzes the state of plasma with the measured conditions instantaneously. According to our estimation, the most time consuming part of MARG2D is eigensolver and we must solve resulted eigensystem whose dimension is hundred thousand within a second. However, current massively parallel processor (MPP) type supercomputer is not applicable for such instantaneous calculation, because the overhead of network communication becomes dominant. Therefore, we employ Cell cluster system, whose processor has higher performance than MPP’s, because we can obtain sufficient processing power with small number of processors. Furthermore, we developed novel eigenvalue solver with the consideration of hierarchical architecture of Cell cluster: Inter processor, intra processor, and SIMD parallelism. Finally, we succeeded to solve the block tridiagonal Hermitian matrix, which had 1024 diagonal blocks and the size of each block was 128 x 128 within a second.

Component-Wise Meshing Approach and Evaluation of Bonding Strategy on the Interface of Components for Assembled Finite Element Analysis of Structures

The finite elements are extensively utilized to solve various problems in engineering fields with the growth of computing technologies. However, there is a lack of methodology for analyses of huge assembled structures. The mechanics on the interface of each components, for instance, contact, bolt joint and welding in assembly is a key issue for important huge structure such as nuclear power plants. On the other hand, it is well known that as finite element models become large and complex, construction of detailed mesh becomes a bottleneck in the CAE procedures. To solve these problems, the authors would like to introduce component-wise meshing approach and bonding strategy on the interface of components. In order to assemble component-wise meshes, the penalty method is introduced not only to constrain the displacements, but also to introduce classical spring connection on the joint interface, although penalty method is claimed that it is not suitable for iterative solver. In this paper, the convergence performance of an iterative solver with penalty method is investigated and the detailed component-wise distributed computation scheme is described with numerical examples.

Development of a High-speed Eigenvalue-solver for Constant Plasma Monitoring on a Cell Cluster System

Abstract We developed a high speed eigenvalue solver that is an essential part of a plasma stability analysis system for fusion reactors on a Cell cluster system. In order to achieve continuous operation of fusion reactors, we must evaluate the state of plasma within the characteristic confinement time of the plasma density and temperature in fusion reactors. This is because we can prevent plasma from being disrupted by controlling the confining magnetic field, if we can determine the state of the plasma within the characteristic confinement time. Therefore, we introduced a Cell processor that has high computational power and high performance/cost, in order to achieve constant monitoring of fusion reactors. Furthermore, we developed a novel eigenvalue solver, which usually consumes most of the plasma evaluation time, to achieve high performance of our Cell cluster system. The eigensolver is based on the conjugate gradient (CG) method and was designed by considering three levels of parallelism, which we refer to as Intra-processor, Inner-processor, and SIMD parallel. In addition, we developed a new CG acceleration method, called locally complete LU. This method has the same acceleration performance as complete LU, which is one of the best acceleration methods, without any reduction in parallel performance. Finally, we succeeded in obtaining our target performance: we were able to solve a block tri-diagonal Hermitian matrix containing 1024 diagonal blocks, where the size of each block was 128 × 128, within a second. Therefore, we have found a suitable candidate for achieving a satisfactory monitoring system.

High Performance Computing for Analyzing PB-Scale Data in Nuclear Experiments and Simulations

By performance improvement of computers and expansion of experiment facilities, output data having became huge. In near future, the output data will become petabyte (PB)-scale. It will become increasingly important how huge data is analyzed efficiently and derive useful information. To analysis huge data efficiently, we are constructing large-scale data integrated analysis system which treats terabytes-petabytes data. In this system, two elemental technologies, i.e., heterogeneous processor and distributed parallel computing framework with fault-tolerance are implemented. The former and the latter are effective for computation dominant processes and data I/O dominant processes, respectively. First, we have applied acceleration by the heterogeneous processor to experimental data and estimated its performance. The processor accelerated experimental data processing substantially. Next, then we have constructed a prototype of distributed parallel computing system for simulation data and carried out processing test. We have found the notice points for application these elemental techniques.

Development of Cognitive Methodology based Data Analysis System

Nuclear engineering is an integrated engineering field of mechanical and civil engineering, partical physics as well as fluid and thermodynamics. Researchers in nuclear engineering fields need to treat extensive physical and engineering information obtained through theories, simulations, experiments and observations in order to promote a nuclear technology safely and securely. To meet the need, the Cognitive methodology-based Data Analysis System (CDAS) which equips information technologies that have recognition abilities similar to those of humans has been developed. The system supports researchers to analyze numerical simulation data by using extensive scientific knowledge. In the present study, information technology is developed for performing these processes and for configuring systems. In addition, a prototype system has been constructed using this information technology and an application experiment using a virtual plant vibration simulator has been performed to confirm the implementability of the system. The results obtained demonstrate that the CDAS enables researchers to dynamically set essential functions for evaluation and judgment, enabling them to readily extract meaningful and reliable information from large-scale data of up to 1 TB.

Interoperation between Atomic Energy Grid Infrastructure (AEGIS) and Other Grids

Coordination of global knowledge is needed to advance the computational and computer science needed for nuclear research. We have been conducting cooperative international research in various fields to construct a highly-secure worldwide network computing infrastructure, based on the Atomic Energy Grid Infrastructure (AEGIS). A promising way to achieve this is to establish interoperation using AEGIS with other grids. The operation of existing grid environments that allow the continuous development and execution of user applications is critical. To achieve the interoperability while maintaining operations, we have developed a system that converts messages among different grid middlewares without requiring modification of grid middlewares. To realize interoperability with two or more grid middlewares in the present study, we have defined the application programming interface (API) as a common interface to convert messages among the grid middlewares. We have applied our system to three interoperable environments. Through these experiments, we have confirmed that our system is applicable to the construction of interoperable environments among various grid middlewares.