Similar documents
Developing Predictive Capability for High Performance Steady State Plasmas

The National Transport Code Collaboration

Parallel Analysis and Visualization on Cray Compute Node Linux

Support and Development for Remote Collaborations in Fusion Research

Diagnostics. Electric probes. Instituto de Plasmas e Fusão Nuclear Instituto Superior Técnico Lisbon, Portugal

DATA MANAGEMENT, CODE DEPLOYMENT, AND SCIENTIFIC VISUALLIZATION TO ENHANCE SCIENTIFIC DISCOVERY IN FUSION RESEARCH THROUGH ADVANCED COMPUTING

Applications to Computational Financial and GPU Computing. May 16th. Dr. Daniel Egloff

Dynamic Load Balancing of Parallel Monte Carlo Transport Calculations

Major Conclusions of the MFE Study

MEng, BSc Computer Science with Artificial Intelligence

MEng, BSc Applied Computer Science

HPC Deployment of OpenFOAM in an Industrial Setting

Load Balancing on a Non-dedicated Heterogeneous Network of Workstations

Service Oriented Architecture 1 COMPILED BY BJ

Software Development around a Millisecond

GenericServ, a Generic Server for Web Application Development

Efficiency Considerations of PERL and Python in Distributed Processing

On One Approach to Scientific CAD/CAE Software Developing Process

Neptune. A Domain Specific Language for Deploying HPC Software on Cloud Platforms. Chris Bunch Navraj Chohan Chandra Krintz Khawaja Shams

PIE. Internal Structure

Evaluation of fuelling requirements and transient density behaviour in ITER reference operational scenarios

ME6130 An introduction to CFD 1-1

Low Level. Software. Solution. extensions to handle. coarse grained task. compilers with. Data parallel. parallelism.

The Ultra-scale Visualization Climate Data Analysis Tools (UV-CDAT): A Vision for Large-Scale Climate Data

Data Management in an International Data Grid Project. Timur Chabuk 04/09/2007

Enterprise Application Integration

Effective Java Programming. efficient software development

A Survey Study on Monitoring Service for Grid

PyCompArch: Python-Based Modules for Exploring Computer Architecture Concepts

EAI OVERVIEW OF ENTERPRISE APPLICATION INTEGRATION CONCEPTS AND ARCHITECTURES. Enterprise Application Integration. Peter R. Egli INDIGOO.


Portfolio of Products. Integrated Engineering Environment. Overview

GEDAE TM - A Graphical Programming and Autocode Generation Tool for Signal Processor Applications

Data Analysis with MATLAB The MathWorks, Inc. 1

Fronting Integrated Scientific Web Applications: Design Features and Benefits for Regulatory Environments

Panasas High Performance Storage Powers the First Petaflop Supercomputer at Los Alamos National Laboratory

Facilities Perspectives

ICME Platform Foundational Capabilities - Sreedhar Reddy

Advanced spatial discretizations in the B2.5 plasma fluid code

Part IV. Conclusions

GRAVE: An Interactive Geometry Construction and Visualization Software System for the TORT Radiation Transport Code

What can DDS do for You? Learn how dynamic publish-subscribe messaging can improve the flexibility and scalability of your applications.

THE CCLRC DATA PORTAL

Service Oriented Architecture (SOA) An Introduction

SYSTEMS AND SOFTWARE REQUIREMENTS SPECIFICATION (SSRS) TEMPLATE. Version A.4, January 2014 FOREWORD DOCUMENT CONVENTIONS

Programming models for heterogeneous computing. Manuel Ujaldón Nvidia CUDA Fellow and A/Prof. Computer Architecture Department University of Malaga

CHAPTER 1: OPERATING SYSTEM FUNDAMENTALS

PHP FRAMEWORK FOR DATABASE MANAGEMENT BASED ON MVC PATTERN

Data Management/Visualization on the Grid at PPPL. Scott A. Klasky Stephane Ethier Ravi Samtaney

STATISTICA Solutions for Financial Risk Management Management and Validated Compliance Solutions for the Banking Industry (Basel II)

Structural Health Monitoring Tools (SHMTools)

GA A23745 STATUS OF THE LINUX PC CLUSTER FOR BETWEEN-PULSE DATA ANALYSES AT DIII D

Information integration platform for CIMS. Chan, FTS; Zhang, J; Lau, HCW; Ning, A

Tritium Gas Processing for Magnetic Fusion

UPS battery remote monitoring system in cloud computing

PERFORMANCE MONITORING OF JAVA COMPONENT-ORIENTED DISTRIBUTED APPLICATIONS

GA A25827 EFFECTS OF ELECTRON CYCLOTRON CURRENT DRIVE, COUNTER-NBI, AND ROTATIONAL ENTRAINMENT ON NEOCLASSICAL TEARING MODE CONTROL IN DIII-D

FUTURE VIEWS OF FIELD DATA COLLECTION IN STATISTICAL SURVEYS

Software Quality Factors OOA, OOD, and OOP Object-oriented techniques enhance key external and internal software quality factors, e.g., 1. External (v

ParFUM: A Parallel Framework for Unstructured Meshes. Aaron Becker, Isaac Dooley, Terry Wilmarth, Sayantan Chakravorty Charm++ Workshop 2008

Performance Improvement of Application on the K computer

Harnessing the power of advanced analytics with IBM Netezza

Impact of the plasma response in threedimensional edge plasma transport modeling for RMP ELM control at ITER

2 (18) - SOFTWARE ARCHITECTURE Service Oriented Architecture - Sven Arne Andreasson - Computer Science and Engineering.

Pentaho High-Performance Big Data Reference Configurations using Cisco Unified Computing System

Architectural Patterns. Layers: Pattern. Architectural Pattern Examples. Layer 3. Component 3.1. Layer 2. Component 2.1 Component 2.2.

Service-oriented architecture in e-commerce applications

Centralized Systems. A Centralized Computer System. Chapter 18: Database System Architectures

Patterns in Software Engineering

PSS E. High-Performance Transmission Planning Application for the Power Industry. Answers for energy.

Real-Time Analytics on Large Datasets: Predictive Models for Online Targeted Advertising

Distributed Objects and Components

The Service Revolution software engineering without programming languages

Technology to Control Hybrid Computer Systems

Service Oriented Architecture (SOA) Implementation Framework for Satellite Mission Control System Software Design

A new web based elearning Platform for Building Simulation

Integration of C++ digital processing libraries and VTK through Tcl/Tk dynamic loadable extensions

SOFT 437. Software Performance Analysis. Ch 5:Web Applications and Other Distributed Systems

The Mantid Project. The challenges of delivering flexible HPC for novice end users. Nicholas Draper SOS18

PERFORMANCE COMPARISON OF COMMON OBJECT REQUEST BROKER ARCHITECTURE(CORBA) VS JAVA MESSAGING SERVICE(JMS) BY TEAM SCALABLE

Design of Remote data acquisition system based on Internet of Things

Masters in Human Computer Interaction

GridSolve: : A Seamless Bridge Between the Standard Programming Interfaces and Remote Resources

Software design (Cont.)

1z0-102 Q&A. DEMO Version

Distributed Systems Architectures

Layering a computing infrastructure. Middleware. The new infrastructure: middleware. Spanning layer. Middleware objectives. The new infrastructure

Migrating Legacy Software Systems to CORBA based Distributed Environments through an Automatic Wrapper Generation Technique

Poster Either Oral or Poster Will not attend. Predictive community computational tools for virtual plasma science experiments.

Manage Software Development in LabVIEW with Professional Tools

Masters in Networks and Distributed Systems

Design of Scalable, Parallel-Computing Software Development Tool

A GENERAL PURPOSE DATA ANALYSIS MONITORING SYSTEM WITH CASE STUDIES FROM THE NATIONAL FUSION GRID AND THE DIII D MDSPLUS BETWEEN PULSE ANALYSIS SYSTEM

Data Mining with Hadoop at TACC

Masters in Computing and Information Technology

DAME Astrophysical DAta Mining Mining & & Exploration Exploration GRID

CE 504 Computational Hydrology Computational Environments and Tools Fritz R. Fiedler

Manual for simulation of EB processing. Software ModeRTL

Poisson Equation Solver Parallelisation for Particle-in-Cell Model

Modernizing Simulation Input Generation and Post-Simulation Data Visualization with Eclipse ICE

Transcription:

TRANSPORT CODE NATIONAL (NTCC) COLLABORATION H. Kritz, G. Bateman, M. Erba, J. Kinsey Arnold University Physics Department Lehigh 16 Memorial Drive East, Bethlehem, PA 18015 St. John H. Atomics, San Diego, CA General Cohen, R. Jong, L. Lodestro, T. B. Yang R. Livermore National Laboratory, Livermore, CA Lawrence Greenwood, W. Houlberg D. Ridge National Laboratory, Oak Ridge, TN Oak McCune, D. Mikkelsen, A. Pletzer D. Plasma Physics Laboratory, Princeton, NJ Princeton R. Cary, K. G. Luetkemeyer J. Corporation, Boulder, CO Tech-X Wiley J. of Texas, Austin, TX University APS Centenial Meeting, Atlanta, GA, 22 March 1999

Address major physics issues facing the fusion program coupling of core and edge physics physics of internal and edge transport barriers self consistent treatment: rotation, transport and MHD investigate both the common and unique physics Design new transport codes with interface that allows remote use Web-invocable OBJECTIVES NTCC the way fusion modeling codes are constructed and used Change Flexible transport codes based on modern software engineering of various toroidal connement devices transport codes to be used by experimentalists, and by modelers theoreticians

COMPONENTS NTCC Library Module Web-based community-owned library of modules Using modern object-oriented computer techniques PYTHON, CORBA, JAVA) (C++, Designed to generate transport codes that are to maintain, customizable, Web-invocable, user friendly easy Most modules extracted from existing FORTRAN codes Reviewed to ensure standards Available at http://w3.pppl.gov/ntcc See poster GP01.92 Framework

Web-invokable applications using the NTCC framework address physics issues facing the fusion program to Customizable for dierent physics applications See poster GP01.93 Organize workshops on modern computing techniques the fusion research community for Demonstration Codes Available at http://electrojet.colorado.edu/wwwntcc/ Education

In contrast to specialized computations which explore and large scale instabilities in isolation turbulence Integrated modeling can bring strongly interacting physics in context together THE NEED FOR TRANSPORT CODES to make coherent, self-consistent predictions about fusion devices experimental to explore synergies between turbulent transport, conditions (wall-plasma interactions), boundary and abrupt rearrangements (large scale instabilities) to test theoretical predictions against experimental data to understand the physics of systematic trends as the scaling with, q, and ) (such

anomalous transport driven by turbulence neoclassical and parallel transport radiative transport auxiliary heating, fueling, momentum and current drive atomic physics nuclear reactions equilibrium force balance large scale instabilities plasma-wall interactions COMPONENTS OF TRANSPORT CODES

PHYSICS ISSUES NEW TRANSPORT CODES FOR study physics of internal and edge transport barriers predict plasma rotation and compare with measurements test non-local transport simulate self-organized cirticallity test multiple transport models against experimental data examine trends in parameter scans self-consistent treatment of rotation, transport, and MHD study coupling of core and edge physics

STRUCTURE OF CLIENT-SERVER DEMONSTRATION CODES NTCC NTCC Demo Code can use up to 3 computers simultaneously: Client provides graphical user interface user runs client on local computer physics and data server may run on a remote computer data may be stored on a third computer controls physics and/or data server can produce plots input and results written in JAVA for Web access

Physics server advances the transport equations Data server accesses experimental data CORBA-IDL is an industry-wide standard used for network communications computes transport form physics modules will compute sources and sinks numerical schemes to advance transport equations written in PYTHON, C++, FORTRAN options to access ITER Prole Database or MDS+ database

USE FIVE COMPUTER WHY LANGUAGES? Each language is designed to handle a dierent task best CORBA-IDL (Interface Denition Language) is an network communication language object-oriented modules can be switched in or out on the y (C++, FORTRAN, PYTHON, CORBA, JAVA) JAVA for client provides a rich set of Web-invocable User Interface (GUI) components on any computer Graphical CORBA (Common Object Request Broker Interface) the way data transmitted from an application standardizes on one computer to an application on another computer PYTHON is an object-oriented scripting language used rapid development and control of applications for

A C++ wrapper is written for each FORTRAN module standardize link to other languages to The object-oriented languages are designed to work together PYTHON, CORBA, and JAVA) (C++, Part of the NTCC objective is to experiment with new and computational techniques languages FORTRAN used in modules extracted from legacy codes C++ is an ecient high-performance object-oriented language standard

WITH USING FIVE PROBLEMS COMPUTER LANGUAGES DIFFERENT There is much more to learn Each language has to be installed on each developer's computer There is no standard coupling with FORTRAN is dierent from one platform to the next) (linking Compilers are dierent on dierent platforms Not all the compilers are up to the most recent standard example, C++ was standardized in November 1997, For not all C++ compilers have Standard Template Library but and Namespaces... which are part of the new standard

Portability variety of machine architectures and operating require little commercial software with any commercial systems; Modularity code composed of independent modules which be documented, tested, and executed in isolation; graphics can Web/network awareness extraction of data from experimental remote access to module documentation, test databases; Web-based control and remote rendering write client server applications to allow invocation of the framework and SOFTWARE GUIDELINES software isolated and with a general license obtained packages separated cases, source code; couple in modules over the network the Internet; the client application steer the application over render graphical output of the results and

Distributed design framework modules running in processes - as a design tool even if the main computation dierent Use embeddable and extendible object-oriented scripting facilitates rapid prototyping of framework class language Salablity demonstrate signicantly new and powerful way building software compared to current practice; stimulate of ultimately resides in a single address space and algorithms, provides easy extension of functionality structures of compiled code modules, enables customization the product without aecting archived compiled code, of facilitates re-use of modules in dierent combinations and for numerical experiments development and renewed interest in testing theories further models and

Scalability accommodate a large range of complexity the modules; be able to accommodate coupling to edge in Inheritance interfaces should allow for inheritance to layered complexity, with the lower layers simple and have Reuse of legacy code The framework will be separate existing transport codes, but will reuse previously written from ranging from a simple point model boundary condition models a full 2-D (or even 3-D) edge code, and coupling to to neutrals models ranging from simple to full Monte Carlo reusable. to the maximum possible extent; invoke the modules code the module library from

DIRECTIONS FOR NTCC FUTURE PHYSICS direct coupling between turbulence simulations and plasma proles macroscopic incorporation of large-scale MHD phenomena their interactions with the microscale turbulence and three-dimensional computations of neutral atoms from the wall and from beam heating neutral interaction with material wall needed to determine erosion rates, recycling, plasma contamination and determine the self-consistent eects of eddy currents, plasma-coil interactions, neutron radiation and of complexity and widely diering scales poses a unique Mix challenge for transport codes. For example: computational coupling core and edge and wall complex atomic physics and radiation transport

PROBLEMS PARTICULAR TO THE SOME EDGE PLASMA There are narrow spatial scales coupled with large scales, Strong interaction between atomic physics and plasma physics Rapid changes caused by Edge Localized Modes Turbulence scale length comparable to gradient scale length such as: H-mode pedestal radiating boundary layer in detached divertor plasma sheath at material boundary rapid transitions from long to short mean free path

Existing transport codes combine models for sources, sinks, eects of large scale instabilities, neutrals,... transport, Many of these models entail analytic approximations to physical phenomena complex More sophisticated models are needed for: MODELING CORE PLASMA turbulence driven transport accurate modeling of transport barriers eects of RF heating (e.g., momentum deposition) frequency of sawtooth oscillations and ELMs

DIRECTIONS FOR NTCC FUTURE COMPUTING Need to enhance parallel computational eciency of major such as components, Integrated modeling requires development of software using algorithms for dierent components disparate Parallelization requirements are dierent at dierent times, dierent physical processes, as well as for dierent regions for 3-D turbulence, nonlinear MHD instabilities, and neutral source models Monte-Carlo of plasma

PARALLELISM and DATA PARALLELISM FUNCTIONAL e.g., Compute dierent sources and sinks in parallel Integrated modeling requires heterogeneous parallelization Data Parallelism same computation performed many 1. in parallel times e.g., Monte Carlo computation may following each A atom on a dierent processor sample Functional Parallelism to launch dierent kinds of 2. on dierent processors in parallel computations data parallism (domain decomposition) and parallism (process decomposition) functional

Use data parallism and functional parallism together in an problem dependent manner optimal, Control these computations with adaptive load leveling in which the program automatically migrates or grid points where they are needed most inserts HETEROGENEOUS PARALLELIZATION short processes automatically grouped together long processes automatically divided up into granules smaller optimum granularity may change with time and be dierent for dierent simulations may this process is logically similar to an adaptive grid

HETEROGENEOUS DEVELOPING PARALLELIZATION Use object-oriented programming to dene computational of variable granularity see objects Some components themselves require massive parallelization Develop adaptive granularity load-leveling algorithms Develop algorithms to couple components Construct framework with coupling algorithms and adaptive to integrate components granularity http://www.tc.cornell.edu/edu/talks/introtopp/less.html parallelization is a relatively new approach Heterogeneous high performance computing. in

Funding for NTCC began in April 1998 More than 10 modules have been submitted to the Library (see poster GP01.92) Module NTCC framework has been developed Web-invocable demonstration codes have been developed poster GP01.93) (see STATUS OF NTCC http://p3.pppl.gov/ntcc Using C++, PYTHON, FORTRAN, CORBA, and JAVA http://electrojet.colorado.edu/wwwntcc/ Physics and data servers JAVA based client