The Center for High Performance Computing. Dell Breakfast Events 20 th June 2016 Happy Sithole

Similar documents
HPC Progress and Response to the National Cyber-Infrastructure

Support for Large Scale HPC and Big Data Science Projects in South Africa: SKA and More. 25 th September 2017

The Stampede is Coming Welcome to Stampede Introductory Training. Dan Stanzione Texas Advanced Computing Center

C Y B E R - I N F R A S T R U C T U R E I N S O U T H A F R I C A. Daniel Adams Portfolio Committee 3 June 2015

Governor Patrick Announces Funding to Launch Massachusetts Open Cloud Project Celebrates Release of 2014 Mass Big Data Report

The Cambridge Bio-Medical-Cloud An OpenStack platform for medical analytics and biomedical research

CSD3 The Cambridge Service for Data Driven Discovery. A New National HPC Service for Data Intensive science

N a t i o n a l I C T R & D a n d I n n o v a t i o n R o a d m a p

JÜLICH SUPERCOMPUTING CENTRE Site Introduction Michael Stephan Forschungszentrum Jülich

NERSC Site Update. National Energy Research Scientific Computing Center Lawrence Berkeley National Laboratory. Richard Gerber

HPC Innovation Lab Update. Dell EMC HPC Community Meeting 3/28/2017

Preparing GPU-Accelerated Applications for the Summit Supercomputer

HPC projects. Grischa Bolls

The Stampede is Coming: A New Petascale Resource for the Open Science Community

INSPUR and HPC Innovation. Dong Qi (Forrest) Oversea PM

Inauguration Cartesius June 14, 2013

InfiniBand Strengthens Leadership as the Interconnect Of Choice By Providing Best Return on Investment. TOP500 Supercomputers, June 2014

WVU RESEARCH COMPUTING INTRODUCTION. Introduction to WVU s Research Computing Services

Umeå University

Umeå University

Organizational Update: December 2015

The Impact of Inter-node Latency versus Intra-node Latency on HPC Applications The 23 rd IASTED International Conference on PDCS 2011

Data Life Cycle. Research. Access Collaborate. Acquire. Analyse. Comprehend. Plan. Manage Archive. Publish Reuse

CYFRONET SITE REPORT IMPROVING SLURM USABILITY AND MONITORING. M. Pawlik, J. Budzowski, L. Flis, P. Lasoń, M. Magryś

Building the Ecosystem for ARM Servers

Mission-Critical Lustre at Santos. Adam Fox, Lustre User Group 2016

Altair RADIOSS Performance Benchmark and Profiling. May 2013

Jülich Supercomputing Centre

INSPUR and HPC Innovation

Computer Science Section. Computational and Information Systems Laboratory National Center for Atmospheric Research

Emerging Technologies for HPC Storage

PART-I (B) (TECHNICAL SPECIFICATIONS & COMPLIANCE SHEET) Supply and installation of High Performance Computing System

CS500 SMARTER CLUSTER SUPERCOMPUTERS

HPC and AI Solution Overview. Garima Kochhar HPC and AI Innovation Lab

Users and utilization of CERIT-SC infrastructure

HYCOM Performance Benchmark and Profiling

Comet Virtualization Code & Design Sprint

Intel Enterprise Edition Lustre (IEEL-2.3) [DNE-1 enabled] on Dell MD Storage

TECHNICAL GUIDELINES FOR APPLICANTS TO PRACE 13 th CALL (T ier-0)

SPARC 2 Consultations January-February 2016

Monash High Performance Computing

University at Buffalo Center for Computational Research

Interconnect Your Future

Introduction to National Supercomputing Centre in Guangzhou and Opportunities for International Collaboration

The Cray CX1 puts massive power and flexibility right where you need it in your workgroup

HPC IN EUROPE. Organisation of public HPC resources

NAMD Performance Benchmark and Profiling. January 2015

Update on LRZ Leibniz Supercomputing Centre of the Bavarian Academy of Sciences and Humanities. 2 Oct 2018 Prof. Dr. Dieter Kranzlmüller

Greek e-infrastructures Short report

GRIDS INTRODUCTION TO GRID INFRASTRUCTURES. Fabrizio Gagliardi

NVIDIA Update and Directions on GPU Acceleration for Earth System Models

LAMMPS-KOKKOS Performance Benchmark and Profiling. September 2015

Accelerating Insights In the Technical Computing Transformation

The Spider Center-Wide File System

Planning for Liquid Cooling Patrick McGinn Product Manager, Rack DCLC

How то Use HPC Resources Efficiently by a Message Oriented Framework.

NCAR Workload Analysis on Yellowstone. March 2015 V5.0

InfraStructure for the European Network for Earth System modelling. From «IS-ENES» to IS-ENES2

INCREASE IT EFFICIENCY, REDUCE OPERATING COSTS AND DEPLOY ANYWHERE

Building an Exotic HPC Ecosystem at The University of Tulsa

Himeno Performance Benchmark and Profiling. December 2010

System Design of Kepler Based HPC Solutions. Saeed Iqbal, Shawn Gao and Kevin Tubbs HPC Global Solutions Engineering.

For personal use only

HPC Capabilities at Research Intensive Universities

Habanero Operating Committee. January

Sentinel-1A Data Dissemination and Users Data Uptake. Jolyon Martin (ESA) Betlem Rosich Tell (ESA) Michael Foumelis (RSAC c/o ESA)

2014 LENOVO INTERNAL. ALL RIGHTS RESERVED.

Update on Cray Activities in the Earth Sciences

Towards Evidence-based Policy in Africa: ICT Access and Usage in 17 African countries.

AN INTRODUCTION TO CLUSTER COMPUTING

Smarter Clusters from the Supercomputer Experts

Part 2: Computing and Networking Capacity (for research and instructional activities)

BlueGene/L. Computer Science, University of Warwick. Source: IBM

Your Roadmap to POWER9: Migration Scenarios

Description of the European Big Data Hackathon 2019

HPC Current Development in Indonesia. Dr. Bens Pardamean Bina Nusantara University Indonesia

CLOUDS OF JINR, UNIVERSITY OF SOFIA AND INRNE JOIN TOGETHER

Netweb Technologies Delivers India s Fastest Hybrid Supercomputer with Breakthrough Performance

Making Supercomputing More Available and Accessible Windows HPC Server 2008 R2 Beta 2 Microsoft High Performance Computing April, 2010

Overview of Tianhe-2

The Energy Challenge in HPC

Basic Specification of Oakforest-PACS

UCX: An Open Source Framework for HPC Network APIs and Beyond

IT Town Hall Meeting

E-Strategies in Africa

IBM Power Advanced Compute (AC) AC922 Server

Technology Changes in Data Centers

Performance Boost for Seismic Processing with right IT infrastructure. Vsevolod Shabad CEO and founder +7 (985)

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

Oncilla - a Managed GAS Runtime for Accelerating Data Warehousing Queries

DELL EMC VALIDATED SYSTEM FOR VIRTUALIZATION

IBM TotalStorage FAStT900 Storage Server New Offering Expands IBM SAN Storage Solutions

VxRail: Level Up with New Capabilities and Powers GLOBAL SPONSORS

Accelerating Computational Science and Engineering with Heterogeneous Computing in Louisiana

e-infrastructures in FP7: Call 7 (WP 2010)

I/O Monitoring at JSC, SIONlib & Resiliency

CLOUD ECONOMICS: HOW TO QUANTIFY THE BENEFITS OF MOVING TO THE CLOUD

CP2K Performance Benchmark and Profiling. April 2011

Dell EMC Ready Bundle for HPC Digital Manufacturing Dassault Systѐmes Simulia Abaqus Performance

HPC Saudi Jeffrey A. Nichols Associate Laboratory Director Computing and Computational Sciences. Presented to: March 14, 2017

Transcription:

The Center for High Performance Computing Dell Breakfast Events 20 th June 2016 Happy Sithole

Background: The CHPC in SA CHPC

User Community: South Africa CHPC Existing Users Future Users

Introduction of CHPC CHPC is a national facility within the NICIS providing HPC resources and services to academia and industry. A broad range of Science and Engineering domains use HPC. The Center is funded by the Department of Science and Technology. Building capacity within the country and continent as part of the SKA project.

Vision An accomplished and preferred partner for HPC solutions in the country and the region Mission Provide world-class HPC that enable cuttingedge research with high impact on the South African economy

CHPC New Infrastructure System Configuration Phase 1 Dell PowerEdge C6320 Servers: Standard Compute nodes 128GB memory/ node 1 008 2 x Intel Xeon E5-2690 v3, 2.6 GHz (Haswell) processors (12 Cores Each 24 cores / node) Dell PowerEdge R930 servers: 24 192 Large Memory Compute Nodes 1024GB / node (FAT nodes) 5 4 x Intel Xeon E7-4850 v3 processors, 2.2 GHz (14 Cores Each 56 cores / node) Infiniband FDR 2:1 non-blocking (56 Gbps) Parallel Storage (Useable) PB 4 Total Number of Racks (including Compute, Login, Management and Storage Nodes) Centos 7.1 Measured Linpack Compute (Tflop/s) 782.9 280 19

LENGAU Launch

Lustre Parallel File System@CHPC Intel Lustre 2.5.4 - Intel Management for Lustre (IML) - 4PB useable IOR Results

Future Upgrades Phase 2 upgrade to be completed by end of October 2016. Initial power restriction removed, now 2 MWatts of power available to the data center. Cooling configuration for the additional servers could be considered (for example on DLC or immersion) Processor Roadmap (including accelerators) coupled with Dell Server Roadmap will assist the technology decision for the upgrade. Interconnect fabric scalable to 1PFLOP from initial design. Storage expansion to be considered in conjunction with archiving options. Optimisation of fan speed on the servers is necessary.

Fluent F1 racecar

WRF-3.7, Intel vs gcc

CASTEP Tsessebe Lengau Before optimization Tsessebe Initialisation time = 6.33 s Calculation time = 1033114.91 s Finalisation time = 72.00 s Total time = 1033193.24 s Peak Memory Use = 710728 kb it took approximately 286 hours ~ 11 days. Overall parallel efficiency rating: Terrible (16%) After optimization Email: pngoepe@ul.ac.za after optimization Lengau Initialisation time = 2.58 s Calculation time = 64277.31 s Finalisation time = 9.00 s Total time = 64288.88 s Peak Memory Use = 986236 kb It took approximately 18 hours =3/4 day. Overall parallel efficiency rating: Very good (82%)

60 km grid spacing 8 km grid spacing Ultra high-resolution climate modelling on the CHPC clusters 1 km grid spacing Computational resources determine grid spacing Subgrid-processes parameterised Parameterisations source of uncertainty

Extending CSIR climate modelling to Africa on the CHPC clusters Tailor-made projections for Mozambique, Ethiopia, Egypt, Nigeria and Namibia currently running on new CHPC cluster CCAM applied in stretchedgrid mode 8 km simulations nudged in lower resolution CORDEX simulations 1 km resolution simulations over of 200 x 200 km^2 nudged within 8 km res simulations C160 stretched-grid with 8 km horizontal resolution over a domain of about 1500 x 1500 km^2, centered over Port Said

Simulation period: 2009-2012 Multiple nudging: ERA reanalyses 75km to 8km (1300kmx1300km domain size) 1km (150kmx150km)

Industry Initiatives What are the Drivers? Highly scalable methods for modelling and simulation exploit massive parallelism and data visualisation. Heterogeneous computing environments require new programming models.. Emerging real-time, data-intensive applications require robust HPC capability. Energy requirements increase industry overheads. Tax incentive models from government. SASOL De Beers Johnson Matthey ESKOM XTRATA ECJIV NNR TOTAL MINTEK HATCH Virtual Prototypes Offerings Applications are made to scale. HPC is accessible to businesses and applications of all types. HPC is effectively utilised. Easier to migrate applications to other HPC systems.

African Data Intensive Research Cloud

Human Capital Development Initiatives Annual Winter School targeting new post-graduate researchers and focus on basics of HPC and induction into HPC facilities, hosted in conjunction with Universities around the country. Annual Introductory Scientific Programming School focusing on parallel programming and python packaged for post-graduate students. Domain focused workshops arranged in collaboration with research communities and tailored to meet the needs of the researchers. New architecture, such as the heterogeneous computing environments (GP-GPU and Xeon-Phi). System Administrator training with hands-on configuration on HPC system. Providing administrators with smaller HPC systems that they manage at their own institutions.

Student Cluster Challenge The Student Cluster Challenge is aimed at training students in HPC. The program is focused at undergraduate students, from all S&E disciplines. National selection process starts in June (Winter School) and provides teams that compete in December. Normally 10 teams of 4 students compete. The winning team in the national competition + 2 best from runner-ups form a team that represent South Africa in the International competition. The team spends a week at Dell Research Laboratory in Austin in January for further training. Dell is the major sponsor with hardware and logistics for both local competition and international team. Mellanox provide the interconnect. The students are recruited to different HPC entities in the country (For example, CHPC, SKA etc..)

The CHPC Team Wins!! (Added post ISC)

Noteworthy Projects Dell ICT Academy announced in 2015 through equity equivalence. Building already completed and the first intake of students (10 high school graduates and 5 graduates). Some of the graduates from the academy will form Dell Channel partners with strong focus on HPC system integration. Focus is on enterprise system and high performance computing. Discussions to include a team from Academy in SC program. CHPC Solution Center to focus on testing of new technologies with Dell in conjunction with other solutions centers such as Cambridge University and PISA.