Cyber-infrastructure developments in South Africa: Enhancing Industrial

Competitiveness

By Dr. Happy Sithole, CSIR-NICIS

Slide #2

Background

Slide #3

NICIS Model

• Common services between anytwo entities of NICIS

• Joint planning is very important

• Intersection of the three entitiesmost crucial, as defines thesuccess of overall NICIS.

• Development of cross-cuttingskills internally and within thecommunity

SANReN DIRISA

CHPC

HPDAFast/Secure systems

Transferring large data

Mainly large Scale Science projects SKA, CERN, Bioinformatics, 4IR, SADC C.I. Framework, Climate Change

Slide #4

Vision

The realisation of a vibrant and competitive knowledge basedeconomy impacting socio-economic development by enablingeducation, research, and innovation through shared access toadvanced cyberinfrastructure facilities and services

Slide #5

Mission

To provide a world class national integrated cyberinfrastructure system thatenables research, innovation and learning comprising a national highperformance computing facility, a national research and education networkand a national data intensive research infrastructure accessible across theresearch and higher education sector through integrated eResearch servicesand the development of relevant human capital.

Slide #6

• Open (FAIR) Data & Open Science • Federated locally and globally

(“One-stop-shop” catalogue)• Certified as Trusted Repository• Linked to funder systems• Suite of services for RDM and

data intensive analytics

40 PB

2 PB

Archival data & staging: VM access

8 PB

Active data: near real time interactive access

0.5 PB

Services & staging between DIRISA and CHPC

storage systems

Storage Virtualisation

Server

Fast Parallel FS

1.6 PFLOPSComput

e6 PB

Software defined storage hierarchySmall, fast Big, slow

iRODS

DIRISA cloud portal

DIRISA Architecture

Slide #7

NREN Roll-Out

Slide #8

Definitions and Concepts

The Fourth Industrial Revolution aims to leverage differences between the physical, digital, and

biological sphere. It integrates cyber-physical systems and the Internet of Things, big data and

HPC, robotics, artificial-intelligence based systems and additive manufacturing.

https://ec.europa.eu/digital-single-market/en/cyber-physical-systemshttps://ec.europa.eu/digital-single-market/en/internet-thingshttps://ec.europa.eu/digital-single-market/en/big-datahttps://ec.europa.eu/digital-single-market/en/robotics

Slide #9

FIR Technologies

The fourth industrial revolution is driven by technological advancement, new

and emerging technologies and platforms. Some of the technologies to be

embraced in the era of the fourth industrial revolution:

• Artificial intelligence and machine learning;

•Advanced robotocs and new forms of automation;

•Ubiquitous mobile internet;

•Sensors and the internet of things;

•Blockchain and distributed ledgers;

•3D priniting;

•New matterials;

•Genetic advances, bio-engineerng; and

•Quantum computing.

Slide #10

Digital skills for youth: Opportunities & challenges

Robots Internet of Things

AI/ML/DL

High Performance Data Analytics .e.g DIRISA

Cloud computingHPC

Slide #11

Under threat New jobs

Slide #12

Are we preparing for the workplace of the future?

Slide #13

Why HPC?

Experiments and Prototypes

Virtual Prototypes

HPCCompetitiveness

IDEAProduct to Market

Out-Compete to Out-Compute

Slide #14

Design It

Build It

Test It

Fix It!

Slide #15

Technology Development Continuum

Theoretical&

BasicResearch

Commercialization&

Production(.com or .org)

AppliedPrototyping

&Development

Optimization&

Robustification

HPCBridges the Gap

BETWEENBasic Research & Commercialization

Product Life Cycle

Phase 1Feasibility

Phase 3Prototyping

Phase 4Production/Deployment

Phase 0Concept/

Vision

Universities& Labs

PrivateIndustry

Application

Phase 2Design/

Development

Over the horizon development …

Slide #16

Pillars of HPC Ecosystem

Sources include: Heidelberg Collaboratory, Minsker, GMES

Integrated, Information-Based Decision Making

Visualization Dashboard

Information-Based Decisions

Calibration

Automated Provenance

Human, Physical, Virtual Sensors

…Seamless Computing…

Archives

Data Fusion

ModelingMultimodal

Slide #17

• Uptake of simulation goes beyond traditional Science and Engineering due to the “Big Data Analytics”.

• High fidelity models required for “Virtual prototyping” improve with compute intensity.

• Availability of high quality data and more processing power is enabler.

• Strengths in both domain knowledge and HPC are differentiator.

CHPC

Prototypes

Institutional/Mid-Range Clusters

Application

Desktops

Model Development

Imp

rove

d a

ccu

racy

An

sw

ers

, F

aste

r an

d A

ccu

rate

High Performance Data Analytics Trends

Slide #18

Basics of HPC

Slide #19

Lengau HPC System

• 496 on TOP500

• Awarded the fastest supercomputer in Continent

Slide #20

❑ Graphical Processor Unit (GPU) Cluster: 24 V100’s

❑ 6 nodes (36 Intel Gold 6150 CPU’s; 3 V100’s; PCIe)

❑ 3 nodes (40 Intel Gold 6150 CPU’s; 4 V100’s; NVLink)

❑ Since September 2018

❑ Immediate Demand from users:

❑ Transfer of Chemistry Users (MD)

❑ Resources for Machine Learning (ML)

❑ Machine Learning applications optimised – TensorFlow❑ Big Data Analytics Focused

Machine Learning and AI Applications

Slide #21

% C

ore

s in

Use

# o

f Jo

bs

Running

Queued

19 Jun 2019 18 Jul 2019

HPC Utilisation

Slide #22

# Research Programmes

210 Active Programmes

# Core Hours

97 million hours

SA Academic72%

SA Public Sector14%

SA Industry5%

African SKA9%

SA Academic80%

SA Public Sector12%

SA Industry1%

African SKA7%

Resource Utilisation

19 Jan – 18 Jul 2019

Slide #23

Building aerodynamics(ECI-JV)

Radio-Astronomy Dish Design(University of Stellenbosch)

High-speed Train Design (UP):

Elutriator Design (De Beers Marine)

Computational Mechanics / Engineering

Centrifugal Steam Compressor(Aerotherm)

Slide #24

❑ DC Furnace Plasma Arc electric smelting simulations

❑ CFD and Magnetohydrodynamics (MHD) simulations

❑ Resulted in patent for Mintek on arc detection technology

❑ Fully dependent on CHPC for HPC resource requirments

Mineral Processing Applications

Cross-section of DC FurnaceShowing Plasma Arc (photo) SimulatedPhotographic Image

Slide #25

eMendi Building Completed in June 2017 at

Port of Ngqura – https://www.youtube.com/watch?v=EGItgLSO-y4

ECI-JV: Architecture

Computational Mechanics / Engineering

CASE STUDY

Building aerodynamics (Eastern Cape Infrastructure-JV)

https://www.youtube.com/watch?v=EGItgLSO-y4

Slide #26

Gas reservoir simulation – Eclipse Software Compositional Benchmark at CHPC

Sbusiso Mkhize (PetroSA)

Natural Gas Extraction - PetroSA

Slide #27

Materials ScienceNanoscale Battery technology Simulations at University of Limpopo (UL)

CASE STUDY

❑ CHPC-MMC (Materials Modelling Centre) Collaboration on DLPOLY Scaling

❑ 717 703 atom Li11Mn13O32 nano-particle➔ 112 nodes (2688 cores) Scaling

❑ 13 000 times speedup achieved on CHPC resources

Courtesy: Dr Anton Lopis (CHPC)

Slide #28

Drug Discovery

[3] Scientific Reports, 2015, 5, 11539, doi: 10.1038/srep11539

Slide #29

Dr Thomas Franz and Dr Malebogo Ngoepe (UCT)

CFD Modelling of

thrombus

development in

Cerebral

Aneurisms…

CHPC Staff Support:

Dr Charles Crosby

Computational Mechanics / Engineering

Slide #30

• To develop Mouth Dissolving Film (slim thickness) Formulation containing anti-tubercular drugs (e.g. pyrazinamide, isoniazid etc.) for the effective treatment of TB

Health – TB Drug DevelopmentDr Oluwatoyin Adeleke – Sephako Makgatho University

Slide #31

CHPC-SAWS MoU:

❑ CHPC is Fail-Over Facility

❑ All SAWS Research done at CHPC

South African Weather Service (SAWS)

1st UM forecast which ran operationally on the CHPC in April 2018 (Dr S Landman)

Direct Collaboration with InstitutionsWhat are CHPC Engagement Models?

EXAMPLE

Slide #32

Three stage program:

(1) Winter school: 22 teams

(2) CHPC National Meeting: 10 Teams

(3) ISC: 1 Team representing SA

ISC’15 Runner-Up

Frankfurt

ISC’14 Champions

Leipzig

ISC’17 2nd Place

ISC’19 1st Place

ISC’18 3rd Place

Student Cluster Competition (SCC)

Leipzig Frankfurt

ISC’13 Champions ISC’16 Champions

Human Capital Development and Training