MIKE by DHI 2014 e sviluppi futuri



Similar documents
MIKE 21 Flow Model FM. Parallelisation using GPU. Benchmarking report

High Performance Computing in CST STUDIO SUITE

MIKE BY DHI SAAS PORTAL. MIKE by DHI Software as a Service (SaaS) Step-by-step guide

Flood Modelling for Cities using Cloud Computing FINAL REPORT. Vassilis Glenis, Vedrana Kutija, Stephen McGough, Simon Woodman, Chris Kilsby

HPC Wales Skills Academy Course Catalogue 2015

SOFTWARE FOR WATER ENVIRONMENTS

Part I Courses Syllabus

PRIMERGY server-based High Performance Computing solutions

10- High Performance Compu5ng

GPUs for Scientific Computing

LSKA 2010 Survey Report Job Scheduler

A Theory of the Spatial Computational Domain

Enterprise HPC & Cloud Computing for Engineering Simulation. Barbara Hutchings Director, Strategic Partnerships ANSYS, Inc.

High Performance. CAEA elearning Series. Jonathan G. Dudley, Ph.D. 06/09/ CAE Associates

CUDA programming on NVIDIA GPUs

STATE OF NEVADA Department of Administration Division of Human Resource Management CLASS SPECIFICATION

HPC Deployment of OpenFOAM in an Industrial Setting

Multi-core Curriculum Development at Georgia Tech: Experience and Future Steps

Cloud Computing. Alex Crawford Ben Johnstone

Arcane/ArcGeoSim, a software framework for geosciences simulation

P013 INTRODUCING A NEW GENERATION OF RESERVOIR SIMULATION SOFTWARE

Microsoft Compute Clusters in High Performance Technical Computing. Björn Tromsdorf, HPC Product Manager, Microsoft Corporation

22S:295 Seminar in Applied Statistics High Performance Computing in Statistics

SGI HPC Systems Help Fuel Manufacturing Rebirth

IBM Platform Computing : infrastructure management for HPC solutions on OpenPOWER Jing Li, Software Development Manager IBM

HPC enabling of OpenFOAM R for CFD applications

GPU System Architecture. Alan Gray EPCC The University of Edinburgh

Introduction to GPU hardware and to CUDA

Data Centric Systems (DCS)

Local Area Networks: Software

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

Cost Savings Solutions for Year 5 True Ups

Multicore Parallel Computing with OpenMP

Unleashing the Performance Potential of GPUs for Atmospheric Dynamic Solvers

Best practices for efficient HPC performance with large models

Building an Internal Cloud that is ready for the external Cloud

Obj: Sec 1.0, to describe the relationship between hardware and software HW: Read p.2 9. Do Now: Name 3 parts of the computer.

Cellular Computing on a Linux Cluster

CS 3530 Operating Systems. L02 OS Intro Part 1 Dr. Ken Hoganson

Compute Cluster Server Lab 3: Debugging the parallel MPI programs in Microsoft Visual Studio 2005

The GPU Accelerated Data Center. Marc Hamilton, August 27, 2015

Clusters: Mainstream Technology for CAE

LBM BASED FLOW SIMULATION USING GPU COMPUTING PROCESSOR

SURFsara HPC Cloud Workshop

NVIDIA CUDA Software and GPU Parallel Computing Architecture. David B. Kirk, Chief Scientist

Recent Advances in HPC for Structural Mechanics Simulations

HPC Cluster Decisions and ANSYS Configuration Best Practices. Diana Collier Lead Systems Support Specialist Houston UGM May 2014

High Performance Computing

Using the Windows Cluster

Parallels Server 4 Bare Metal

Relations with ISV and Open Source. Stephane Requena GENCI

HPC Software Requirements to Support an HPC Cluster Supercomputer

A general-purpose virtualization service for HPC on cloud computing: an application to GPUs

Scaling Objectivity Database Performance with Panasas Scale-Out NAS Storage

1 Bull, 2011 Bull Extreme Computing

OpenPOWER Outlook AXEL KOEHLER SR. SOLUTION ARCHITECT HPC

Maximize Performance and Scalability of RADIOSS* Structural Analysis Software on Intel Xeon Processor E7 v2 Family-Based Platforms

Cross Platform Mobile. -Vinod Doshi

Emerging Technology for the Next Decade

Scalable Data Analysis in R. Lee E. Edlefsen Chief Scientist UserR! 2011

MEng, BSc Applied Computer Science

Basin simulation for complex geological settings

Cloud Computing and Amazon Web Services

Journée Mésochallenges 2015 SysFera and ROMEO Make Large-Scale CFD Simulations Only 3 Clicks Away

VMware Horizon DaaS: Desktop as a Cloud Service (DaaS)

Lecture 1 Introduction to Parallel Programming

Comparing the performance of the Landmark Nexus reservoir simulator on HP servers

Content Distribution Management

~ Greetings from WSU CAPPLab ~

Automating Big Data Benchmarking for Different Architectures with ALOJA

Scalable and High Performance Computing for Big Data Analytics in Understanding the Human Dynamics in the Mobile Age

The Construction of Seismic and Geological Studies' Cloud Platform Using Desktop Cloud Visualization Technology

Cluster, Grid, Cloud Concepts

Interoperability between Sun Grid Engine and the Windows Compute Cluster

Applicata Plans & prices

Agenda. HPC Software Stack. HPC Post-Processing Visualization. Case Study National Scientific Center. European HPC Benchmark Center Montpellier PSSC

Week Overview. Installing Linux Linux on your Desktop Virtualization Basic Linux system administration

MEng, BSc Computer Science with Artificial Intelligence

Maximizer CRM 12 Summer 2013 system requirements

Performance Analysis of a Numerical Weather Prediction Application in Microsoft Azure

PyCompArch: Python-Based Modules for Exploring Computer Architecture Concepts

Performance And Scalability In Oracle9i And SQL Server 2000

So#ware Tools and Techniques for HPC, Clouds, and Server- Class SoCs Ron Brightwell

Planning Your Installation or Upgrade

Efficient Load Balancing using VM Migration by QEMU-KVM

Recommended hardware system configurations for ANSYS users

Introduction to HPC Workshop. Center for e-research

Robust Algorithms for Current Deposition and Dynamic Load-balancing in a GPU Particle-in-Cell Code

Transcription:

MIKE by DHI 2014 e sviluppi futuri Johan Hartnack Torino, 9-10 Ottobre 2013

Technology drivers/trends Smart devices Cloud computing Services vs. Products

Technology drivers/trends Multiprocessor hardware OS Stored data

Overview Performance - Parallelization Remote execution - Utilizing common hardware (cloud) Data - Gaining access to relevant data MIKE HYDRO - Next generation water ressources products User involvement - How to influence the process DHI

MIKE by DHI 2014 e sviluppi futuri Multiprocessors - Performance DHI

Performance ~ Parallelization Shared memory Distributed memory Graphical processing unit DHI

Parallelization Shared memory (OPENMP): The calculations are carried out on multiple processors on the same pc all accessing the same memory.

Parallelization MIKE 21 Single domain, hydrodynamic calc. Speedup 2 cores: 15-30% 4 cores : 40-80% Excl. Side-feeding Incl. Side-feeding

Parallelization Distributed memory The calculations are carried out on multiple processors each with its own memory space and required information is passed between the processors at regular intervals

Basic concept Message passing interface (MPI) standard interface used for communication between processors The distribution of work is based on the domain decomposition concept (physical sub-domains) Each processor integrates basic equation in sub-domain Data exchange between sub-domains is based on halo layer/elements concept I/O is handled on local level

Basic concept

High Performance Computing Distributed memory Optimisation and benchmarking Example of the results of test of parallelisation on a 864 core Linux cluster

HPC - an investment High performance computing (HPC) has been one of the fastest growing ITmarkets within the last five years Date Linux Unix Mixed MS Windows BSD based June 2013 95.2% 3.2% 0.8% 0.6% 0.2%. DHI

Utilizing the GPU for numerics Fairly cheap to purchase Get a speed up factor at a cheap rate NVIDIA based cards DHI

Parallelization - GPU GPU

Parallelization - GPU GPU(Graphical Processing Unit): The main calculations are carried out on the GPU processors. Data are transferred as needed MIKE 21 FM based Only HD part

Parallelization - GPU Benchmark Mediterranean sea Not possible to scale the degree of parallelization Scale using the resolution of the mesh DHI

Benchmark preliminary results double precision DHI

Benchmark preliminary results single precision DHI

Preliminary indications Performance dependent on GPU hardware Good scalability DHI

MIKE by DHI 2014 e sviluppi futuri Remote simulation service(cloud) DHI

Remote Simulation 42 A 42 B DHI 30 October, 2013 #22

Remote Simulation How it works: When your model simulation is ready to run, simply activate the new simulation console, select the executing computer and launch the simulation Your simulation is then executed on this remote computer and the result files are easily transferred to your PC when the simulation has ended It is possible to run as many simulations in parallel using your remote computer resources as your licence allows This also means that you can use remote simulation with AUTOCAL for automatic model calibration / optimization DHI 2012

Remote Simulation How to get started: Remote Simulation Console DHI 2012

Remote Simulation Availability : Available for MIKE Zero and MIKE URBAN based products from release 2014 Available for Corporate and Subscription type licenses Your simulation resources are limited by your hardware and by your MIKE by DHI licence (the number of cores and the number of simultaneous runs ) DHI 2012

MIKE by DHI 2014 e sviluppi futuri DHI WaterData Data service DHI

A new service from DHI Making knowledge about water environments accessible Water knowledge Software and tools Tailored solutions and DSS Knowledge sharing Data fit for use Consultancy MIKE by DHI MIKE CUSTOMISED by DHI THE ACADEMY by DHI DHI Free MIKE SMA Subscribe Buy Publically available data, licence restrictions Entry level product Global coverage Processed and ready to use Handling fees but semi automated Medium processing Derived products High value products Additional value and services

DHI

DHI

MIKE by DHI 2014 e sviluppi futuri MIKE HYDRO next generation water resources modelling DHI

MIKE HYDRO MIKE HYDRO introduced in Release 2012. The vision: Common platform for (most) Water Resources products Overall features: Map centric, easy-to-use Graphical User Interface Usability and work-flow oriented design No third party GIS components required MIKE Zero component One setup-editor DHI #31

MIKE HYDRO River The Graphical User Interface River model tree view items River model toolbar icons Cross sections plot Graphical River Network editor Structures plot DHI #32

MIKE HYDRO Release 2014 includes: MIKE HYDRO River, Phase I River modelling with MIKE HYDRO First release of classic MIKE 11 GUI successor Includes a subset of classic MIKE 11 GUI features DHI #33

MIKE HYDRO Basin Water Quality using ECO Lab ECO Lab: Numerical lab for Ecological modelling Open and Generic ECO Lab tool in for MIKE customized HYDRO: water quality models Utilizes ECO - Lab eliminates Templates hard-coded with mathematical WQ formulas descriptions of ecosystems Templates are - increased open and editable flexibility - enhanced usability MIKE HYDRO Basin; WQ editor DHI #34

MIKE by DHI 2014 e sviluppi futuri MIKE User council How to influence the MIKE by DHI path DHI

MIKE User Council The primary mission of the MIKE by DHI User Council (in short: MIKE UC) is to provide input to DHI in improving the MIKE products so that they cover the most important modelling needs as seen from the perspective of the members of the MIKE UC. The vision of the MIKE UC is to be able to see tangible improvements in each new release of the MIKE products based on their input. DHI 2012

User ideas now part of release Tool for describing dikes in MIKE 21 s topography DHI 2012

User ideas now part of release Water balance tool for MIKE FLOOD DHI 2012

User ideas now part of release MIKE URBAN Gridded Rainfall DHI 2012

Summary Multiprocessors - Variety of options (MIKE 21 FM GPU) Cloud computing - Remote execution and SaaS Service - DHI WaterData Next gen. MIKE - MIKE HYDRO User involvement - MIKE User council DHI 2012

Thank you Johan Hartnack Torino, 9-10 Ottobre 2013 DHI