Grid Computing in Multidisciplinary CFD optimization problems Toan NGUYEN May 13-15th, 2003 Project OPALE Parallel CFD Conference, Moscow (RU) The challenge of Multi-physics Industrial Applications

PARALLEL CFD OPTIMIZATION STATE OF THE ART FUTURE TRENDS & CONCLUSION OUTLINE MULTIDISCIPLINARY APPLICATIONS CURRENT ISSUES INRIA

PART 1

Created 1967 French Scientific and Technological Public Institute Ministry of Research and Ministry of Industry INSTITUT NATIONAL DE RECHERCHE EN INFORMATIQUE ET AUTOMATIQUE National Research Institute for Computer Science and Automatic Control

INRIA MISSIONS Fundamental and applied research Design experimental systems Technology transfer to industry Knowledge transfer to academia International scientific collaborations Contribute to international programs Technological assessment Contribute to standards organizations

Rocquencourt Rennes Lorraine Sophia Antipolis Rhône-Alpes in six Research Centers 900 permanent staff 400 researchers 500 engineers, technicians and administrative pers. 500 researchers from other organizations 600 trainees, PhD and post- doctoral students 100 external collaborators 400 visiting researchers from abroad Budget 120 MEuros (tax not incl.) 25% self-funding through 600 contracts Futurs PERSONNEL

CHALLENGES Expertise to program, compute and communicate using the Internet and heterogeneous networks Design new applications using the Web and multimedia databases Expertise to develop robust software Design and master automatic control for complex systems Combine simulation and virtual reality

APPLICATIONS Telecommunications and multimedia Healthcare and biology Engineering Transportation Environment

RESEARCH PROJECTS Teams of approx. 20 researchers Medium-term objectives and work program (4 years) Scientific and financial independence Links and partnerships with scientific and industrial partners on national and international basis Regular assessment of results during given time- scale

PROJECTS 99 Projects in four themes: 1. Networks and Systems 2. Software Engineering and Symbolic Computing 3. Human-Computer Interaction, Image Processing, Data Management, Knowledge Systems 4. Simulation and Optimization of Complex Systems

INTERNATIONAL COOPERATION Develop collaborations with European research centres and industries & strengthen the European scientific community in Information & Communication Technologies Increase international collaborations and enhance exchanges Cooperations with the United States, Japan, Russia Relations with China, India, Brazil, etc. Partnerships with developing countries World Wide Web Consortium (W3C) Work with the best industrial partners worldwide

Areas Located Sophia-Antipolis & Grenoble Follow-up SINUS project INRIA project (January 2002) OPALE NUMERIC OPTIMISATION (genetic, hybrid, …) MODEL REDUCTION (hierarchic, multi-grids, …) INTEGRATION PLATFORMS Coupling, distribution, parallelism, grids, clusters,... APPLICATIONS : aerospace, electromagnetics, …

STATE OF THE ART PART 2

GRID COMPUTING THE GRIDBUS PROJECT (Univ. Melbourne, Australia)

GRID COMPUTING INFORMATION SERVICES RESOURCE MANAGEMENT DATA MANAGEMENT

APPLICATIONS National Partnership for Advanced Computational Infrastructure

GRID COMPUTING HIGH THROUGHPUT COMPUTING HIGH PERFORMANCE COMPUTING PETA-DATA MANAGEMENT LONG DURATION APPLICATIONS

HIGH-PERFORMANCE PROBLEM SOLVING ENVIRONMENTS AFFORDABLE HIGH-PERFORMANCE COMPUTING GRID COMPUTING BUSINESS TO BUSINESS & E-COMMERCE LARGE SCALE SCIENTIFIC APPLICATIONS ENGINEERING, BIO-SCIENCES, EARTH & CLIMATE MODEL. IRREGULAR AND DYNAMIC BEHAVIOR APPLICATIONS

GRID COMPUTING OPTIMALGRID PROJECT (IBM Almaden Resarch Center)

DISTRIBUTED HETERO. DYNAMIC RESOURCES & SERVICES DISCOVERY, SHARING, COORDINATED USE, MONITORING GRID COMPUTING PERFORMANCE, SECURITY, SCALABILITY, ROBUSTNESS DYNAMIC MONITORING ADAPTIVE RESOURCE CONTROL ERROR AMPLIFIER SYNDROM PERFORMANCE DIRECTED MANAGEMENT

BROKERING, FAULT DETECTION & TROUBLESHOOTING GRID COMPUTING PLANNING & ADAPTING DISTRIBUTED APPLICATIONS NEED ENQUIRY, REGISTRATION PROTOCOLS CACHING, MIGRATING, REPLICATING DATA APPLICATIONS : HIGH ENERGY PHYSICS (DATAGRID, PPDG, GriPhyN) GRID SERVICES (OGSA) LOCATION TRANSPARENCY, MULTIPLE PROTOCOL BINDINGS COMPATIBLE UNDERLYING PLATFORMS CREATE & COMPOSE DISTRIBUTED SYSTEMS

GRID COMPUTING NSF Middleware Initiative : Globus, Condor-G, NWS, KX509, GSI-SSH, GPT ISI, Univ. Chicago, NCSA, SDSC, Univ. Wisconsin Madison NSF, Dept Energy, DARPA, NASA GOAL : « national middleware infrastructure to permit seamless resource sharing across virtual organizations » GRID Research, Integration, Deployment & Support center PHILOSOPHY : « the whole is greater than the sum of its parts » APPLICATIONS : NEES, GriPhyN, Intl Virtual Data Grid Lab (ATLAS)

PARALLEL & DISTRIBUTED PROGRAMMING SOFTWARE DEV. : FREE OPEN SOURCE (Linux, FreeBSD) DEVELOPMENT LARGE DISTRIBUTED DATA FILE SYTEMS GRID COMPUTING BEOWULF CLUSTERS HIGH-SPEED GIGABITS/SEC NETWORKS COMPONENT PROGRAMMING Incentives

BEOWULF CLUSTER PC-cluster at INRIA Rhône-Alpes (216 Pentium III procs.)

« Grid is a type of parallel and distributed system that enables the sharing, selection, and aggregation of resources distributed across multiple administrative domains, based on their (resources) availability, capability, performance, cost and users' quality-of-service requirements. If distributed resources happen to be managed by a single, global centralised scheduling system, then it is a cluster. In cluster, all nodes work cooperatively with common goal and objective as the resource allocation is performed by a centralised, global resource manager. In Grids, each node has its own resource manager and allocation policy. » Rajkumar Buyya (Grid Infoware) GRIDS vs. CLUSTERS

PARALLELISM IS NOT DISTRIBUTION DISTRIBUTION SUPPORTS A LIMITED FORM PARALLELISM DISTRIBUTION vs. PARALLELISM PARALLELISM ALLOWS DISTRIBUTION GLOBUS WILL NOT PARALLELIZE YOUR CODE YOU CAN DISTRIBUTE SEQUENTIAL CODES YOU CAN DISTRIBUTE PARALLEL CODES YOU CAN RUN SEQUENTIAL CODES IN « PARALLEL » YOU CAN RUN SEQUENTIALLY PARALLEL CODES

WHERE WE ARE TODAY 1980 : one year CPU time 1992 : one month « » 1997 : four days « » 2002 : one hour « » ASCI White (LLNL) : IBM SP Power 3 procs MCR Linux (LLNL) : Intel 2.4 GHz Xeon procs ASCI Q (LANL) : HP Alpha procs Bits and pieces… Earth Sim (Japan) : NEC procs Moore’s law results…

DISTRIBUTED SIMULATION PLATFORM MULTI-DISCIPLINE PROBLEM SOLVING ENVIRONMENTS HIGH-PERFORMANCE & TRANSPARENT DISTRIBUTION USING CURRENT COMMUNICATION STANDARDS USING CURRENT PROGRAMMING STANDARDS WEB LEVEL USER INTERFACES OPTIMIZED LOAD BALANCING & COMMUNICATION FLOW What is required...

DISTRIBUTED : LAN, WAN, HSN... CODE-COUPLING FOR HETEROGENEOUS SOFTWARE COLLABORATIVE APPLICATIONS COMMON DEFINITION, CONFIGURATION, DEPLOYMENT, EXECUTION & MONITORING ENVIRONMENT TARGET HARDWARE : NOW, COW, PC clusters,... TARGET APPLICATIONS : multidiscipline engineering,... INTEGRATION PLATFORMS Distributed tasks interacting dynamically in controlled and formally provable way What they are...

DISTRIBUTED OBJECTS ARCHITECTURE SOFTWARE COMPONENTS COMPONENTS ARE DISTRIBUTED OBJECTS WRAPPERS AUTOMATICALLY (?) GENERATED COMPONENTS ENCAPSULATE CODES DISTRIBUTED PLUG & PLAY

« CAST » INTEGRATION PLATFORM CAST OPTIMIZERS CORBA SOLVERS ServerWrapper Modules

SOFTWARE COMPONENTS BUSINESS COMPONENTS LEGACY SOFTWARE OBJECT-ORIENTED COMPONENTS DISTRIBUTED OBJECTS COMPONENTS CASUAL METACOMPUTING COMPONENTS ? C++, PACKAGES,... Java RMI, EJB, CCM,...

DISTRIBUTED OBJECTS ARCHITECTURE SOFTWARE CONNECTORS CONNECTORS ARE SYNCHRONISATION CHANNELS SEVERAL PROTOCOLS CONNECTORS = DATA COMMUNICATION CHANNELS - SYNCHRONOUS METHOD INVOCATION - ASYNCHRONOUS EVENT BROADCAST COMPONENTS COMMUNICATE THROUGH SOFTWARE CONNECTORS

NEW APPLICATION METHODOLOGIES // SOFTWARE LIBRARIES : MPI, PVM, SciLab //,... PARALLEL and/or DISTRIBUTED HARDWARE NESTING SEVERAL DEGREES PARALLELISM PARALLEL APPLICATIONS DOMAIN DECOMPOSITION GENETIC ALGORITHMS GAME THEORY HIERARCHIC MULTI-GRIDS The good news….

NESTING PARALLELISM LEVERAGE OPTIMISATION STRATEGIES COMBINE SEVERAL APPROACHES DOMAIN DECOMPOSITION GENETIC ALGORITHMS // SOFTWARE LIBRARIES : MPI,... GRIDS PC-CLUSTERS …

Lays the ground for GRIDS and METACOMPUTING PC & Multiprocs CLUSTERS : thousands GHz procs... HIGH-SPEED NETWORKS : ATM, FIBER OPTICS... ADVANCES IN HARDWARE GLOBUS, LEGION CONDOR, NETSOLVE Gigabits/sec networks available (2.5, 10, …) The best news….

CLUSTER COMPUTING PC-cluster at INRIA Rhône-Alpes (216 Pentium III Itanium procs. Linux)

PARALLEL CFD OPTIMIZATION PART 3

« CAST » INTEGRATION PLATFORM GOALS TESTBED “DECISION” CORBA INTEGRATION PLATFORM DESIGN FUTURE HPCN OPTIMISATION PLATFORMS COLLABORATIVE MULTI-DISCIPLINE OPTIMISATION GENETIC & PARALLEL OPTIMISATION ALGORITHMS CODE COUPLING FOR CFD, CSM SOLVERS & OPTIMISERS C OLLABORATIVE A PPLICATIONS S PECIFICATION T OOL

The front stage….

PROCESS ALGEBRA

TEST CASE SHOCK-WAVE INDUCED DRAG REDUCTION WING PROFILE OPTIMISATION (RAE2822) Euler eqns (Mach 0.84, aoa = 2°) + BCGA (100 gen.) 2D MESH : nodes, triangles 4.5 hours CPU time (SUN Micro SPARC 5, Solaris 2.5) 2.5 minutes CPU time (PC cluster 40 bi-procs, Linux)

TEST CASE WING PROFILE OPTIMISATION

CAST DISTRIBUTED INTEGRATION PLATFORM NICE RENNES GRENOBLE PC cluster n CFD solvers CAST GA optimiser PC cluster software VTHD Gbits/s network

APPLICATION EXAMPLE MULTI-ELEMENT WING PROFILE OPTIMISATION

APPLICATION EXAMPLE WING GEOMETRY

APPLICATION EXAMPLE OPTIMISATION STRATEGY

Cas de test NprocCPU (seconde) Accélération (T1/Ti) APPLICATION EXAMPLE PERFORMANCE DATA 1h 35 mn 6 mn

APPLICATION EXAMPLE PERFORMANCE DATA

APPLICATION EXAMPLE PERFORMANCE DATA

The results...

Check for syntaxe of request NSD ORB MICO Event channell, i1, i2, i3, …. IRD Algogen.idl AlgoGen i1,i2, i3, …, in CAST CfdSolver cfd1 CfdSolver cfd2 « CAST » INTEGRATION PLATFORM Behind the stage, again... GRID 3 PC-CLUSTERS

Event channel, i 1, i 2, i 3, …, i n CfdSolver Cfd1 Processor P0 Processor P1 Processor P3 Processor P2 i1i1 CfdSolver Cfd2 Processor P0 Processor P1 Processor P3 Processor P2 i2i2 CfdSolver Cfd3 Processor P0 Processor P1 Processor P3 Processor P2 i3i3 Genetic Algorithm i 1, i 2,i 3, …, i n Parallelized with MPI on 4 processors CORBA server implemented in C++ CORBA client implemented in C++ EMBEDDED PARALLELISM

APPLICATION EXAMPLE PERFORMANCE DATA

* Curves quasi-parallels => same speed up, whatever the place. * Join an horizontal asymptote: time = 200 s APPLICATION DEPLOYMENT The game : load balancing,...

MULTIDISCIPLINARY APPLICATIONS PART 4

Data Bases Modeling Deterministic/Stochastic Optimizers Validation methods Aeroacoustic s Aerodynamics Aeroelasticity Safety Medical application Drag reduction Industrial multi physics test cases & requirements Database Graphic analysis tools Validation guidelines Noise reduction Electronics facilities Multi-Physics, Numerical Analysis, Applied mathematics, grid computing Thermal flows Aeronautics Propulsion Communication System Integration Platform Pollution reduction Fluid atmospheric environment MULTIDISCIPLINARY APPLICATIONS

HIGH PERFORMANCE COMPUTING HIGH THROUGHPUT COMPUTING APPLICATIONS REQUIREMENTS MULTI-LAYERED ARCHITECTURE HIGH ENERGY PHYSICS CERN LHC FACILITY BIOSCIENCES, ENGINEERING, ENVIRONMENTAL APPS, … SATELLITE IMAGING

SHOULD OR COULD A GRID EMULATE A MAINFRAME ? HOW CAN COMPUTE MODELS BE ADAPTED TO MAKE BEST USE OF GRIDS ? APPLICATIONS REQUIREMENTS WHERE DO GRIDS NOT MAKE SENSE ? WHAT IS THE REAL COST OF OWNING A GRID ? CAN UNUSED POWER OF DESKTOP BE HARNESSED ? HOW TO USE GRIDS FOR HIGH I/O APPLICATIONS ? HOW TO DESIGN GRIDS FOR HIGH AVAILABILITY ?

EXISTING PLATFORMS Globus, Condor, NetSOLVE, Legion, …. DESIGN ALTERNATIVES EXISTING TOOLS NWS, SUN GRID ENG….

DESIGN ALTERNATIVES HARWARE & SOFTWARE ORIENTED ENV. PROBLEM ORIENTED ENVIRONMENTS Optimize specific pbs & solution : ReMAP (Madeleine, DIET, FAST…) System devlpt & optimisation : PARIS (PADICO, PACO, DO…) OASIS (ProActive, …) APACHE (Athapascan, …) APPLICATION ORIENTED Ease of use : OPALE (CAST), …

INTEGRATING MULTIDISCIPLINARY APPLICATIONS INTEGRATION OF PARTNERS’ EXPERTISE TO DEPLOY COLLABORATIVE APPLICATIONS NETWORKED PC-CLUSTERS, COMPUTERS & DATABASES TO SUPPORT MULTIDISCIPLINARY CHALLENGES HIGH-LEVEL PROCEDURES FOR CONCURRENT ENGINEERING (CSCW, VIRTUAL ORGANIZATIONS & ENTERPRISES …) INCLUDE CAD/CAM, MULTI-PHYSICS SOLVERS & OPTIMIZERS

SCALABILITY AIRFOIL OPTIMIZATION ONERA M6 SUPERSONIC WING AOA = 3°, MACH 1.8 OptimizedInitial profile

PLATFORM REQUIREMENTS NEED FOR VIRTUAL REALITY ENVIRONMENT ? NEED FOR CSCW PROCEDURES & SUPPORT ? NEED FOR GRID COMPUTING ? NEED FOR DISTRIBUTED DATABASE TECHNOLOGY ?

PERFORMANCE AIRFOIL OPTIMIZATION ONERA M6 SUPERSONIC WING AOA = 3°, MACH 1.8

MULTIPHYSICS APPLICATIONS New methods and tools ( validation and optimization ) for solving Multidisciplinary Industrial Challenges Multi Physics Validation expertise spread in Research and Industry Cross fertilize Modeling, Experimentation and Scientific disciplines Single expertize revisited in a multi-disciplinary context : Complexity at interfaces: validation of interfaces in multi physics, multi-scale and multi-modeling to provide a unified view of experiments and numerics

ROBUSTNESS

MULTIPHYSICS APPLICATIONS Multidisciplinary/Multicriteria Optimization expertise spread in Research and Industry Complexity of search spaces: robustness and efficiency of hybridized deterministic/adaptive optimization methods - deterministic and global optimizers - evolutionary optimizers - hierarchy, game strategies and decision methods Complexity at interfaces:CAD/CAM and Parameterization/Optimization

NEW CHALLENGES MULTIDISCIPLINARY DESIGN HIGH-LIFT DEVICES : 1 CRITERION / 1 DISCIPLINE (3D Navier-Stokes) : MAXIMIZE LIFT DRAG-BUFFETING : 2 CRITERIA / 1 DISCIPLINE (3D Navier- Stokes) : MINIMIZE CRUISE DRAG & MAXIMIZE Cz BUFFET AERO-ACOUSTICS & HIGH-LIFT DEVICES : 2 CRITERIA/ 2 DISCIPLINES (3D Navier-Stokes) : NOISE REDUCTION OF MULTI- ELEMENTS AIRFOILS DURING TAKE-OFF SUPERSONIC REGIME & BANG : 2 CRITERIA/ 2 DISCIPLINES (3D Navier-Stokes) SUPERSONIC REGIME & NOISE REDUCTION : 2 CRITERIA/ 2 DISCIPLINES (3D Navier-Stokes)

Distributed Data Bases Local Solvers Deterministic/Stochastic Optimizers Validation codes RESEARCH CENTRES AND UNIVERSITIES INDUSTRIES Industrial multi physics test cases High performance computers Local Databases Graphic analysis tools Validation guidelines Multi-Physics optimisation PC clusters GOVERMENTAL INSTITUTIONS Generic multiphysics test cases PC clusters Communication System Web-based system Computing System Grid computing environment Concurrent engineering platform THE PLATFORM

COMMUNICATION SYSTEM Supports interactions among partners and collaborative applications A DISTRIBUTED DATA MGT SYSTEM Supports remote partners data and test-cases A COMPUTING SYSTEM Supports partners grid-computing resources (PC-clusters, files, …)

CURRENT ISSUES PART 5

APPLICATIONS CHARACTERIZATION MULTIDISCIPLINE OPTIMIZATION MULTIDISCIPLINE MODELLING ONGOING EFFORTS ONGOING EFFORTS AERO-STRUCTURE, AERO-ACOUSTICS : tight coupling COMBUSTION, POLLUTION, NOISE REDUCTION DISTRIBUTED APPLICATIONS SCHEDULING I/O PATTERNS, REAL-TIME ADAPTIVE RESOURCE CONTROL DYNAMIC MONITORING loose coupling

COLLABORATIVE PROJECTS Performance monitoring : dynamic load balancing Integrating applications with grid computing technology Dynamic resource co-allocation, process & data migration Virtual organisations ONGOING EFFORTS ONGOING EFFORTS

MAY OVERLAP & SPECIFIC VIEWS FEDERATED RESOURCES DYNAMIC COLLECTIONS USERS & RESOURCES DISTRIBUTED ALLOCATION MANAGEMENT & SCHEDULING VIRTUAL ORGANISATIONS MEMBERSHIP & ACCESS PROTOCOLS SCALABLE & ROBUST ARCHITECTURE & PROTOCOLS AGGREGATIONS OF DISTRIBUTED RESOURCES (VIRTUE)

VIRTUAL ORGANISATIONS

HIERARCHICAL, GLOBALLY UNIQUE NAMES UNRELIABLE FAILURE DETECTORS VIRTUAL ORGANISATIONS RESOURCE NAME + PROVIDER SCOPE & NAME INFORMATION PROVIDER + AGGREGATE DIRECTORIES + VO GRIS : GRID RESOURCE INFORMATION SERVICE (GLOBUS) DISK SPLITTING (PABLO, AUTOPILOT)

GENERIC INFO. SERVICES FOR RESOURCE DISCOVERY VIRTUAL ORGANISATIONS : VIRTUE (Dan Reed, UIUC) DISTRIBUTED APPLICATIONS STEERING (AUTOPILOT) INTEGRATION WITH GRIDS MONITOR EXISTENCE & CHARACTERISTICS RESOURCES SERVICES & COMPUTATIONS MANAGEMENT INTERACTIVE REAL-TIME (I/O ?) PERFORMANCE TUNING

Sensor design PERFORMANCE MONITORING

How to integrate them in new PSE (Fortran, MPI vs. C, Java, C++) ? LEGACY & NEW APPS Interface with PSE (Sockets, CORBA, RMI, EJB, CCM, …) ? Coupling with existing apps & maths libraries (user transparency) ? Last but not least…

FUTURE TRENDS PART 6

DYNAMIC LOAD BALANCING & RESSOURCE ALLOC « COTS » PROGRAMMING METACOMPUTING TOMORROW’S PSE « COMPONENTS OFF THE SHELF » « POWER SUPPLY PARADIGM APPLIED TO COMPUTING RESOURCES WORLDWIDE » Behind the stage, again... MONITOR, START, SUSPEND, RESUME, STOP, MIGRATE REMOTE PROCESSES DYNAMICALLY

CONCLUSION VIRTUAL ORGANIZATIONS « COTS » PROGRAMMING METACOMPUTING FLEXIBLE & INTEROPERABLE APPS DEVELOPMENT LARGE SCALE MULTIDISCIPLINARY APPLICATIONS COLLABORATIVE ENVIRONMENTS REAL CSCW ON FULL SCALE PRODUCTION PROJECTS FULL USER CONTROL

CONCLUSION « THE DIGITAL DYNAMIC AIRCRAFT » LARGE DYNAMIC COLLABORATIVE ENVIRONMENTS

REFERENCES