Cosmological simulations on High Performance Computers

Similar documents
PRACE hardware, software and services. David Henty, EPCC,

Relations with ISV and Open Source. Stephane Requena GENCI

Welcome to the. Jülich Supercomputing Centre. D. Rohe and N. Attig Jülich Supercomputing Centre (JSC), Forschungszentrum Jülich

Supercomputing Resources in BSC, RES and PRACE

How To Build A Supermicro Computer With A 32 Core Power Core (Powerpc) And A 32-Core (Powerpc) (Powerpowerpter) (I386) (Amd) (Microcore) (Supermicro) (

PRACE the European HPC Research Infrastructure. Carlos Mérida-Campos, Advisor of Spanish Member at PRACE Council

International High Performance Computing. Troels Haugbølle Centre for Star and Planet Formation Niels Bohr Institute PRACE User Forum

Information about Pan-European HPC infrastructure PRACE. Vít Vondrák IT4Innovations

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

Access, Documentation and Service Desk. Anupam Karmakar / Application Support Group / Astro Lab

Kriterien für ein PetaFlop System

Extreme Scaling on Energy Efficient SuperMUC

High Performance Computing. Course Notes HPC Fundamentals

How Cineca supports IT

HPC-related R&D in 863 Program


Evoluzione dell Infrastruttura di Calcolo e Data Analytics per la ricerca

Appro Supercomputer Solutions Best Practices Appro 2012 Deployment Successes. Anthony Kenisky, VP of North America Sales

Computational infrastructure for NGS data analysis. José Carbonell Caballero Pablo Escobar

Building a Top500-class Supercomputing Cluster at LNS-BUAP

David Vicente Head of User Support BSC

Extreme Scale Compu0ng at LRZ

BSC - Barcelona Supercomputer Center

PARALLEL & CLUSTER COMPUTING CS 6260 PROFESSOR: ELISE DE DONCKER BY: LINA HUSSEIN

Data Centric Systems (DCS)

JuRoPA. Jülich Research on Petaflop Architecture. One Year on. Hugo R. Falter, COO Lee J Porter, Engineering

Supercomputer Center Management Challenges. Branislav Jansík

Access to the Federal High-Performance Computing-Centers

1 Bull, 2011 Bull Extreme Computing

Mitglied der Helmholtz-Gemeinschaft. System monitoring with LLview and the Parallel Tools Platform

OpenMP Programming on ScaleMP

Unleashing the Performance Potential of GPUs for Atmospheric Dynamic Solvers

HPC Wales Skills Academy Course Catalogue 2015

Jean-Pierre Panziera Teratec 2011

High Performance Computing at CEA

walberla: A software framework for CFD applications on Compute Cores

COMP/CS 605: Intro to Parallel Computing Lecture 01: Parallel Computing Overview (Part 1)

The PRACE Project Applications, Benchmarks and Prototypes. Dr. Peter Michielse (NCF, Netherlands)

Pedraforca: ARM + GPU prototype

Parallel file I/O bottlenecks and solutions

SGI High Performance Computing

White Paper The Numascale Solution: Extreme BIG DATA Computing

Sun Constellation System: The Open Petascale Computing Architecture

How To Compare Amazon Ec2 To A Supercomputer For Scientific Applications

Big Data Visualization on the MIC

numascale White Paper The Numascale Solution: Extreme BIG DATA Computing Hardware Accellerated Data Intensive Computing By: Einar Rustad ABSTRACT

Dr. Raju Namburu Computational Sciences Campaign U.S. Army Research Laboratory. The Nation s Premier Laboratory for Land Forces UNCLASSIFIED

Cluster Scalability of ANSYS FLUENT 12 for a Large Aerodynamics Case on the Darwin Supercomputer

High Performance Computing

Petascale Visualization: Approaches and Initial Results

Designing and Building Applications for Extreme Scale Systems CS598 William Gropp

Energy efficient computing on Embedded and Mobile devices. Nikola Rajovic, Nikola Puzovic, Lluis Vilanova, Carlos Villavieja, Alex Ramirez

Clusters: Mainstream Technology for CAE

Mississippi State University High Performance Computing Collaboratory Brief Overview. Trey Breckenridge Director, HPC

Software services competence in research and development activities at PSNC. Cezary Mazurek PSNC, Poland

LS-DYNA Scalability on Cray Supercomputers. Tin-Ting Zhu, Cray Inc. Jason Wang, Livermore Software Technology Corp.

Simulation Platform Overview

SURFsara HPC Cloud Workshop

High Performance Computing, an Introduction to

Performance of the JMA NWP models on the PC cluster TSUBAME.

Accelerating Simulation & Analysis with Hybrid GPU Parallelization and Cloud Computing

White Paper The Numascale Solution: Affordable BIG DATA Computing

Cloud Computing. Alex Crawford Ben Johnstone

Performance Evaluation of Amazon EC2 for NASA HPC Applications!

HPC Update: Engagement Model

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

A GPU COMPUTING PLATFORM (SAGA) AND A CFD CODE ON GPU FOR AEROSPACE APPLICATIONS

PLGrid Infrastructure Solutions For Computational Chemistry

Overview of HPC Resources at Vanderbilt

Parallel Programming Survey

The PHI solution. Fujitsu Industry Ready Intel XEON-PHI based solution. SC Denver

Performance Characteristics of a Cost-Effective Medium-Sized Beowulf Cluster Supercomputer

Cray Gemini Interconnect. Technical University of Munich Parallel Programming Class of SS14 Denys Sobchyshak

PLGrid Programme: IT Platforms and Domain-Specific Solutions Developed for the National Grid Infrastructure for Polish Science

Trends in High-Performance Computing for Power Grid Applications

Hank Childs, University of Oregon

Transcription:

Cosmological simulations on High Performance Computers Cosmic Web Morphology and Topology Cosmological workshop meeting Warsaw, 12-17 July 2011 Maciej Cytowski Interdisciplinary Centre for Mathematical and Computational Modelling (ICM), University of Warsaw Email: m.cytowski@icm.edu.pl

Topics Introduce POWIEW and PRACE projects Describe the potential collaboration models Show examples of joint work Describe how to access Polish and European HPC infrastructure 2

POWIEW project HPC Infrastructure for Grand Challenges of Science and Engineering Consortium of 3 biggest Polish HPC centers (Warsaw, Poznań, Cracow) Competence centers and working groups for Grand Challenges: Climate modelling Materials sciences Cosmology Visual Analysis Neurobiology An interesting collaboration model 3

POWIEW HPC Infrastructure Massively Parallel Processing (MPP) systems Fat-node systems Versatile Symmetric Multiprocessing systems Hybrid systems (GPUs and FPGA) Selection criteria: large scale and complex scientific computations support for long computations (stable systems) mission critical computations (weather forecasting) application driven specialized architectures 4

IBM Blue Gene/P system (notos) IBM Blue Gene/P 1 rack ~14 Tflop/s High performance dedicated interconnect 4 TB RAM 178 TB storage Installation date: December 2010 5

Example: REBEL model on Blue Gene/P Main objectives of joint work: measure the performance of GADGET-3 code on Blue Gene/P architecture test different BG/P execution modes and their performance (pure MPI, MPI + pthreads) test scalability of computations on Blue Gene/P architecture prepare configuration for the 1024 3 Warsaw Universe simulation parallelize the post-processing application 6

Example: REBEL model on Blue Gene/P Computations does not scale in VN mode (1024+ nodes) Best time on 1024 nodes for SMP mode 7

IBM POWER7 system (boreasz) POWER7 IH 1 rack ~78 Tflop/s ~10.2 TB RAM 153 TB storage Installation: Autumn 2011 1TFlop nodes (4 processors) Very high memory bandwidth High performance dedicated interconnect 8

Examples: POWER7 and cosmology Blue Waters project NCSA (University of Illinois) ENZO code will be ported and optimized on Blue Waters infrastructure by the team lead by Brain O Shea More: http://www.ncsa.illinois.edu/news/stories/dark_days/ Performance measurements and optimization (3x improvement) of the Gadget code on POWER7 architecture 9

More examples: Dark matter and bright minds Collaboration with Computer Science students: Summer Schools Thesis Part-time jobs Tomasz Kłos Gadget-2 optimization on RoadRunner architecture Thesis: Cosmological simulations on hybrid architectures Currently: hired by European Space Agency as software developer Jakub Katarzyński Parallelization of post-processing application for the Gadget-3 results The work will be released and available for public use 10

How to access POWIEW resources and services Computational Grants reviewed by the ICM s scientific board Additional info and help: pomoc@icm.edu.pl Special assistance for cosmological type projects: Maciej Cytowski (m.cytowski@icm.edu.pl) Wojciech A. Hellwing (pchela@icm.edu.pl) 11

PRACE Partnership for Advanced Computing in Europe Vision: Enable and support European global leadership in public and private research and development. High Performance Computing resources and services for public research. European HPC pyramid (ESFRI) Tier-0: 3-6 European Centers for Petaflop Tier-0:? European Centers for Exaflop Tier-1: National Centers Tier-2: Regional/University Centers PRACE POWIEW 12

PRACE Resources 1st PRACE System (JUGENE) BG/P by Gauss Center for Supercomputing at Juelich 294912 CPU cores, 144 TB memory 1 PFlop/s peak performance 825.5 TFlop/s Linpack 2nd PRACE System (2010/2011) GENCI (CURIE) Intel based system by Bull ~90,000 cores 1.6 PFlop/s peak performance 3rd PRACE System 2011 HLRS (HERMIT) 4th PRACE System 2012 LRZ (SUPER-MUC) 5th and 6th Systems 2012/2013 Cineca and BSC 13

How to access PRACE resources and services Project Access: intended for individual researchers and research groups including multi-national research groups has a one year duration calls for Proposals for Project Access are issued twice yearly (May and November) 3 Regular Calls so far Example: Large Scale simulations of Ly-alpha and Ly-break galaxies in the high-z universe: Probing the epoch of reionization. Project leader: Gustavo Yepes, Universidad Autonoma de Madrid, Spain Preparatory Access: intended for resource use required to prepare proposals for Project Access applications for Preparatory Access are accepted at any time Type A: Code Scalability Testing Type B: Code development and optimisation by the applicant Type C: Code development with support from experts 14

Example: ENZO model on JUGENE Internal PRACE Call (similar to Preparatory Access) ICM joined project team as PRACE development staff ENZO model was build and optimized on JUGENE supercomputer INITS code used for creating initial conditions for ENZO runs (part of ENZO distribution) was fully parallelized by ICM Large simulations: 4096^3 particles Follow-up: proposal for 3rd PRACE Regular Call Whitepaper: Large Scale Simulations of the Non-Thermal Universe, Claudio Gheller (CINECA), Graziella Ferini (CINECA), Maciej Cytowski (ICM), Franco Vazza (Jacobs University) (available soon at http://www.prace-ri.eu) 15

Summary We invite you to collaborate with us: Parallelization and optimization New algorithm design and implementation Workflow design and implementation Assistance in getting access to POWIEW and PRACE resources Maciej Cytowski (m.cytowski@icm.edu.pl) Wojciech A. Hellwing (pchela@icm.edu.pl) 16

Links POWIEW project: http://www.wielkiewyzwania.pl PRACE project: http://www.prace-ri.eu Blue Waters project: http://www.ncsa.illinois.edu/bluewaters/ ICM site: http://www.icm.edu.pl 17