2017 CHPC National Conference

3 Dec 2017, 07:00 → 7 Dec 2017, 19:00 Africa/Johannesburg

Velmoré Hotel Estate

Welcome to the 11th CHPC National Conference, to be held at the picturesque Velmoré Hotel Estate, Pretoria.

The theme for this year's gathering is: HPC convergence with novel applications models for better service to research and industry. The theme comes as HPC is gaining more interest in application to non-traditional markets, and largely due to the convergence of HPC to Cloud and the re-emergence of Artificial Intelligence and strengthening of Machine Learning in broader applications.

The comprehensive programme will include national and international contributions as well as contributions from our colleagues in the cyberinfrastructure system: the South African National Research Network (SANReN) and the Data Intensive Research Initiative of South Africa (DIRISA). The 3rd and 7th December will be tutorials and workshops, while the main conference will be 4-6 December. Once again, the South African Development Community (SADC) HPC Collaboration Forum will form part of the conference to discuss the SADC HPC framework and implementation plans.

Our expo zone will be showcasing solutions by leading technology companies and will be the student competition battleground where almost 20 teams from universities across the country will be seeking national honours in the Student Cluster Competition and in the Student Cyber Security Competition - a new addition to the conference. We trust you will find an exciting programme and we look forward to meeting our regular and new delegates.

We look forward to seeing you in Pretoria!

Dr Happy Sithole, Director: CHPC

 

Conference sponsorship proposal Conference_Sponsorship_Agreement_2017.pdf

Workshop/BoF/SIG Proposals application-1-tutorial-workshop.odt application-2-BoF-or-SIG.odt

Sunday, 3 December

08:00 → 09:00 Registration

09:00 → 12:30 Sunday Workshop: Big Data Meets HPC

The convergence of HPC, Big Data, and Deep Learning is the next game-changing business opportunity. Apache Hadoop, Spark, gRPC/TensorFlow, and Memcached are becoming standard building blocks in handling Big Data oriented processing and mining. Recent studies have shown that default designs of these components can not efficiently leverage the features of modern HPC clusters, like Remote Direct Memory Access (RDMA) enabled high-performance interconnects, high-throughput parallel storage systems (e.g. Lustre), Non-Volatile Memory (NVM). In this tutorial, we will provide an in-depth overview of the architecture of Hadoop, Spark, gRPC/TensorFlow, and Memcached.

We will examine the challenges in re-designing networking and I/O components of these middleware with modern interconnects, protocols (such as InfiniBand, RoCE) and storage architectures. Using the publicly available software packages in the High-Performance Big Data project (HiBD, http://hibd.cse.ohio-state.edu), we will provide case studies of the new designs for several Hadoop/Spark/gRPC/TensorFlow/Memcached components and their associated benefits. Through these, we will also examine the interplay between high-performance interconnects, storage (HDD, NVM, and SSD), and multi-core platforms to achieve the best solutions for these components and applications on modern HPC clusters.

We also present in-depth case-studies with modern Deep Learning tools (e.g., Caffe, TensorFlow, BigDL) with RDMA-enabled Hadoop, Spark, and gRPC.

09:00 Big Data Meets HPC: Exploiting HPC Technologies for Accelerating Big Data Processing and Management

The convergence of HPC, Big Data, and Deep Learning is the next game-changing business opportunity. Apache Hadoop, Spark, gRPC/TensorFlow, and Memcached are becoming standard building blocks in handling Big Data oriented processing and mining. Recent studies have shown that default designs of these components can not efficiently leverage the features of modern HPC clusters, like Remote Direct Memory Access (RDMA) enabled high-performance interconnects, high-throughput parallel storage systems (e.g. Lustre), Non-Volatile Memory (NVM). In this tutorial, we will provide an in-depth overview of the architecture of Hadoop, Spark, gRPC/TensorFlow, and Memcached. We will examine the challenges in re-designing networking and I/O components of these middleware with modern interconnects, protocols (such as InfiniBand, RoCE) and storage architectures. Using the publicly available software packages in the High-Performance Big Data project (HiBD, http://hibd.cse.ohio-state.edu), we will provide case studies of the new designs for several Hadoop/Spark/gRPC/TensorFlow/Memcached components and their associated benefits. Through these, we will also examine the interplay between high-performance interconnects, storage (HDD, NVM, and SSD), and multi-core platforms to achieve the best solutions for these components and applications on modern HPC clusters. We also present in-depth case-studies with modern Deep Learning tools (e.g., Caffe, TensorFlow, BigDL) with RDMA-enabled Hadoop, Spark, and gRPC.

Speaker: Dr Xiaoyi Lu (The Ohio State University)

10:30 Break

09:00 → 17:00 Sunday Workshop: Intel AI/Deep Learning

Join us for a full day of technical sessions dedicated to Artificial Intelligence / Deep Learning showing how these technologies can be implemented on existing and upcoming CPUs and how performance can be scaled to any workload!

09:00 Intel AI Workshop

Join us for a full day of technical sessions dedicated to Artificial Intelligence / Deep Learning showing how these technologies can be implemented on existing and upcoming CPUs and how performance can be scaled to any workload!

Speaker: Mr Stephen Blair-Chappell (Bayncore)

10:30 Break

12:30 Lunch

15:00 Break

09:00 → 17:00 Sunday Workshop: Introduction to HPC

This tutorial will provide an entry-level presentation of the basic concepts, knowledge, and skills associated with a strong foundation in modern supercomputing. This introductory treatment is based on a first-year graduate course taught jointly at several universities and currently at Indiana University. Variations of this tutorial have been given at SC11, SC12, ISC13, ISC14, ISC15, ISC16, CHPC16, and ISC17. The tutorial will describe supercomputer architecture including multi-core organization and GPUs, parallel programming models like OpenMP and MPI, simple parallel algorithms, system software and tools for scheduling, debugging, performance monitoring and tuning, and underlying technology trends and future directions of this rapidly changing field. Emphasis will be given to practical information including current generation systems, sources of available software, and links for further reading. This tutorial provides the broadest outreach to people new to the field, students, managers, policy makers, and those needing a refresh in this rapidly advancing domain. Live demonstrations will be presented throughout the tutorial on the Big Red II+ Petaflops computer at Indiana University. This full-day tutorial will include hands-on use for those attendees wishing to engage the learning experience at this depth. Questions will be welcomed throughout the presentation.

Convener: Thomas Sterling (Indiana University)

09:00 Introduction to HPC

This tutorial will provide an entry-level presentation of the basic concepts, knowledge, and skills associated with a strong foundation in modern supercomputing. This introductory treatment is based on a first-year graduate course taught jointly at several universities and currently at Indiana University. Variations of this tutorial have been given at SC11, SC12, ISC13, ISC14, ISC15, ISC16, CHPC16, and ISC17. The tutorial will describe supercomputer architecture including multi-core organization and GPUs, parallel programming models like OpenMP and MPI, simple parallel algorithms, system software and tools for scheduling, debugging, performance monitoring and tuning, and underlying technology trends and future directions of this rapidly changing field. Emphasis will be given to practical information including current generation systems, sources of available software, and links for further reading. This tutorial provides the broadest outreach to people new to the field, students, managers, policy makers, and those needing a refresh in this rapidly advancing domain. Live demonstrations will be presented throughout the tutorial on the Big Red II+ Petaflops computer at Indiana University. This full-day tutorial will include hands-on use for those attendees wishing to engage the learning experience at this depth. Questions will be welcomed throughout the presentation.

Speakers: Matthew Anderson (Indiana University), Thomas Sterling (Indiana University)

10:30 Break

12:30 Lunch

15:00 Break

09:00 → 17:00 Sunday Workshop: Introduction to Schrödinger’s drug discovery tools

Schrödinger is a leading provider of scientific software in the drug design industry. The workshop will give an overview of the drug discovery tools available from Schrödinger. Participants will be introduced to our new graphical interface, Maestro 11, and will work through hands-on examples for tasks like docking, pharmacophore based virtual screening and building homology models. The workshop will also touch on recent improvements on FEP+, our framework to run MD based free energy calculations. It will conclude with a general Q&A session were participants can discuss their own scientific projects. A trial version of our software will be available to participants prior to the workshop. Please bring your own laptop!

Workshop Files Optional (but recommended) download for the Schrodinger Workshop Workshop Material (Files, Structures, Results) for the Schrodinger Workshop

09:00 Introduction to Schrödinger’s drug discovery tools

Schrödinger is a leading provider of scientific software in the drug design industry. The workshop will give an overview of the drug discovery tools available from Schrödinger. Participants will be introduced to our new graphical interface, Maestro 11, and will work through hands-on examples for tasks like docking, pharmacophore based virtual screening and building homology models. The workshop will also touch on recent improvements on FEP+, our framework to run MD based free energy calculations. It will conclude with a general Q&A session were participants can discuss their own scientific projects. A trial version of our software will be available to participants prior to the workshop. Please bring your own laptop!

Speaker: Dr Stephan Ehrlich (Schrodinger GmbH)

Workshop Files Optional (but recommended) download for the Schrodinger Workshop Material (Files, Structures, Results, ..) for the Schrodinger Workshop

10:30 Break

12:30 Lunch

15:00 Break

09:00 → 17:00 Sunday Workshop: System administrator’s tools and techniques for user-facing software

HPC services continue to provide ever increasing resources and many academic disciplines that formerly did not rely on computational resources are becoming interested in utilizing computational resources to analyze their problems. This gives us a strong incentive to provide more assistance to users over a wider range of disciplines which in turn taxes resources at HPC facilities. This leads us to investigate and/or develop the necessary tools to support easy-use and performant environments when running computations on HPC systems. TACC has developed a large number of tools that we can categorize as “user-facing software” that our users run (sometimes without their knowledge) to ensure that their jobs perform well, and give assistance to debugging efforts when jobs fail. We have selected four important tools, and plan to conduct a hands-on set up for administrators and/or provide details on configuration and management of them. They are primarily focused on environment control and performance analysis: Lmod, containerization (Singularity), Tau and REMORA.

One of TACC’s most successful open-source software packages is Lmod – the Lua based version of the “Environment Modules”. It is the first thing all TACC users encounter and learn to use when they login to a TACC system. We discuss and deploy this tool, and demonstrate what system administrators and package maintainers need to know to build/install/maintain an environment module tool. We also explain how users benefit from Lmod, and the simple and advanced use cases.

We often see communities in specific domain sciences build out similar environments that they want imported into HPC systems. They invest many hours in building these environments (same as we do at TACC for the environment that we present to users) and these environments usually end up in containers. Containerization substantially impacts system administrators because they are environments within the main environment, and many of them have substantial security implications. We demonstrate how to make containers available and investigate the typical usage.

High Performance in HPC systems is something administrators, developers and users should view as paramount in the design of applications, and setting up the optimal environment for production runs. The TAU profiler and analysis utility provides data for evaluating performance, similar to the GNU profiler, but has an advanced GUI interface to make analysis easier (and for viewing events). Another package, REMORA, is a resource monitoring tool, also developed at TACC, which allows users to easily collect execution data (memory, numa, i/o, networking and cpu load). The management, best usage practices and synergy between the TAU and REMORA tools will be presented.

Conveners: Dr Milfeld Kent (TACC), Mr Nicholas Thorne (Texas Advanced Computing Center)

09:00 System administrator’s tools and techniques for user-facing software

HPC services continue to provide ever increasing resources and many academic disciplines that formerly did not rely on computational resources are becoming interested in utilizing computational resources to analyze their problems. This gives us a strong incentive to provide more assistance to users over a wider range of disciplines which in turn taxes resources at HPC facilities. This leads us to investigate and/or develop the necessary tools to support easy-use and performant environments when running computations on HPC systems. TACC has developed a large number of tools that we can categorize as “user-facing software” that our users run (sometimes without their knowledge) to ensure that their jobs perform well, and give assistance to debugging efforts when jobs fail. We have selected four important tools, and plan to conduct a hands-on set up for administrators and/or provide details on configuration and management of them. They are primarily focused on environment control and performance analysis: Lmod, containerization (Singularity), Tau and REMORA. One of TACC’s most successful open-source software packages is Lmod – the Lua based version of the “Environment Modules”. It is the first thing all TACC users encounter and learn to use when they login to a TACC system. We discuss and deploy this tool, and demonstrate what system administrators and package maintainers need to know to build/install/maintain an environment module tool. We also explain how users benefit from Lmod, and the simple and advanced use cases. We often see communities in specific domain sciences build out similar environments that they want imported into HPC systems. They invest many hours in building these environments (same as we do at TACC for the environment that we present to users) and these environments usually end up in containers. Containerization substantially impacts system administrators because they are environments within the main environment, and many of them have substantial security implications. We demonstrate how to make containers available and investigate the typical usage. High Performance in HPC systems is something administrators, developers and users should view as paramount in the design of applications, and setting up the optimal environment for production runs. The TAU profiler and analysis utility provides data for evaluating performance, similar to the GNU profiler, but has an advanced GUI interface to make analysis easier (and for viewing events). Another package, REMORA, is a resource monitoring tool, also developed at TACC, which allows users to easily collect execution data (memory, numa, i/o, networking and cpu load). The management, best usage practices and synergy between the TAU and REMORA tools will be presented.

Speakers: Dr Milfeld Kent (TACC), Mr Nicholas Thorne (Texas Advanced Computing Center)

10:30 Break

12:30 Lunch

15:00 Break

11:00 → 12:00 Student Cyber Security Challenge: Induction

The Student Cyber Security Challenge is an information security student competition with an emphasis on network security.

Students will have to solve network security problems and identify security issues such as Decrypting Passwords, Geo-locating pictures, Securing Websites, Finding information from TCP traffic and Extracting weak security keys.

Finally, the students will compete in on a live hacking scenario where they have to defend their network infrastructure while attacking their competitors' infrastructure.

Convener: Dr Renier van Heerden (SANReN, CSIR)

12:30 → 13:30 Lunch

13:30 → 17:00 Sunday Workshop: Parallel visualization on Lengau with Paraview and VisIt

This workshop will touch on the following topics:

Overview of the 3D visualization codes Paraview and VisIt
Similarities and differences between Paraview & VisIt
Limitations of single core processing of large data sets
OpenGL, VirtualGL and Mesa
Different approaches to remote visualization - VNC or X-forwarding
Different approaches to parallelisation
Decomposed datasets
Tutorial illustrating remote parallel visualization with Paraview
Tutorial illustrating remote parallel visualization with VisIt
Introductory remarks on the concept of in-situ visualization
Concluding remarks

Convener: Dr Charles Crosby (CHPC)

13:30 Parallel visualization on Lengau with Paraview and VisIt

This workshop will touch on the following topics: 1. Overview of the 3D visualization codes Paraview and VisIt 2. Similarities and differences between Paraview & VisIt 3. Limitations of single core processing of large data sets 4. OpenGL, VirtualGL and Mesa 5. Different approaches to remote visualization - VNC or X-forwarding 6. Different approaches to parallelisation 7. Decomposed datasets 8. Tutorial illustrating remote parallel visualization with Paraview 9. Tutorial illustrating remote parallel visualization with VisIt 10. Introductory remarks on the concept of in-situ visualization 11. Concluding remarks

Speaker: Dr Charles Crosby (CHPC)

15:00 Break

17:00 → 17:10 End of Day 1

End of the first day of Workshops

Monday, 4 December

08:00 → 09:00 Registration

09:00 → 10:30 Plenary: Opening

Convener: Dr Werner Janse Van Rensburg (CHPC)

09:00 Welcome

09:05 Representative of the Mayor's office

09:15 Speech by the Director-General, Department of Science & Technology

Speaker: Dr Phil Mjwara (DST)

09:30 Opening of the 11th CHPC Conference

Speaker: Dr Happy Sithole (CHPC)

09:45 Transforming Artificial Intelligence into Business Advantage

Artificial Intelligence is the “Next Big Thing.” Artificial Intelligence (AI) is the most profound innovation by humans, with a far deeper and wider impact on our lives, than all preceding innovations from steam engine of 1712 to the Internet. AI has the potential to transform the very role of humans in our society. Its disruptive power is evidenced in a growing number of applications even today – Apple’s Siri, Amazon Echo, Google Assistant, IBM Watson and Tesla cars. What is AI? What can it do (and not do)? How does it work? Where can it be applied? This presentation is designed to demystify AI and help your company to harness its unique power to smarten and uplift your business. Using successful business cases the presentation highlights how and where AI is being used. It provides a roadmap for companies starting AI projects with approaches for discovering innovation opportunities that add business and customer value with AI. Finally, it offers a structured approach with “AI&U Canvas” for developing a successful business model with innovative AI ideas. The presentation will touch upon radical transformations using AI, with its newly acquired magic of Machine Learning, for industries as varied as Automotive, Finance, Manufacturing, Retail, Healthcare, Tax and Legal Consultation, Security and Agriculture. Similarly, almost all horizontal business functions like Customer Service, Human Resources, Procurement, Production, Sales and Marketing benefit enormously from AI. Jeff Bezos, CEO and Founder of Amazon says, “Machine Learning will improve every business, every government organization, every philanthropy … basically, there is not an institution in the world that cannot be improved with Machine Learning.” The use of AI for future competitiveness is a must. As a result, it is extremely important for everyone, especially for business decision makers, to understand the power, significance, and relevance that AI will have not only on our lives but also on business and society as a whole. AI is inevitable. The time to capitalize on AI is now.

Speaker: Sharad Gandhi (Market.Consult)

10:30 → 11:00 Break

Refreshment break

11:00 → 12:30 Chemistry: I Chemistry & Health

Convener: Dr Andre Stander (University of Pretoria)

11:00 Computational Modeling to support the development of biosensing devices

Speaker: Prof. Krishna Bisetty (Department of Chemistry, Durban University of Technology, Durban)

11:30 Exploring chemical space using in silico studies to identify novel inhibitors of Acetylcholinesterase, a target for Alzheimer’s disease

Alzheimer’s disease (AD) is the leading cause of dementia.[1] The number of AD cases is increasing at an alarming rate, which is mostly the result of increases in both the population and the average life expectancy due to advances in modern medicine. AD has been shown to be twice as prevalent in African-Americans, calling it the “Silent Epidemic of Alzheimer’s Disease”.[2] This suggests that the disease is particularly relevant to South Africa and Africa due to its demographics. As such, there is a dire need for the development of effective treatments of AD. Acetylcholinesterase (AChE) is a protein that has been identified for the treatment of AD, with four AChE inhibitors having been approved for treatment by the FDA. These inhibitors have significant side-effects and/or short-term bio-availability. It is suggested that AChE inhibitors can act as dual inhibitors by both inhibiting AChE and prevent Amyloid Beta (Aβ) aggregation also associated with AD. The binding pocket of AChE is ~20 Å deep gorge which is highly solvated. It has been shown that the majority of the side-chain residues are highly flexible and this flexibility should be considered when screening is performed. In addition, the entrance of the binding pocket known as the Peripheral Anionic Site (PAS) contributes to Aβ aggregation. Using High Through-put Virtual Screening (HTVS), potential inhibitors of AChE from the BioFocus library of 20’000 housed at the CSIR were identify by using ensemble docking. Ensemble docking involves using multiple conformations of the same protein to take the flexibility of the active site into account. In doing so, an improved enrichment in the results is obtained where fewer false negatives occur and alternative interactions are also considered. Furthermore, ensemble docking scales linearly with increasing number of receptor conformations, whereas if screening of compounds was performed with the active site being considered flexible the computational resources would increase exponentially with increasing number of flexible residues. Significant validation testing was initially done on the HTVS model to check that an acceptable enrichment was obtained by testing the model against known active compounds and decoys. The majority of the receptor structures were obtained from the Protein Database (PDB). To improve the model further, more receptor structures were generated by performing Inducted Fit Docking (IFD) with compounds that were false negatives during the initial HTVS model. Promising hits identified from the BioFocus library will be submitted for bioassay to confirm the inhibition against AChE and Aβ aggregation. References (1) Anand, R.; Gill, K. D.; Mahdi, A. A. Neuropharmacology 2014, 76 (PART A), 27. (2) Alzheimer’s Association. Alzheimer’s Dement. 2014, 10 (2), 1.

Speaker: Mr Carl Johan van der Westhuizen (University of Pretoria)

11:50 Probing the catalytic mechanism of an antifibrotic copper metallodrug

Angiotensin-1-converting enzyme (ACE) is a zinc metalloprotease consisting of two domains with distinct inhibitor binding affinities despite their 90% active site identity. While the C-domain controls blood pressure, the N-domain is selective for cleaving the antifibrotic N-acetyl-Ser–Asp–Lys–Pro peptide. Selective N-domain inhibition thus shows potential for the treatment of fibrosis. Contrary to conventional competitive ACE inhibitors such as Lisinopril, catalytic metallodrugs irreversibly inactivate ACE through the oxidation of active site residues. Since the metallodrug is recycled from the irreversibly inactivated target, it is effective at sub-stoichiometric concentrations and less likely to cause side-effects/toxicity. The aim of this study was to elucidate the molecular mechanism responsible for N-selective ACE inactivation by Cu-Gly-Gly-His-Lisinopril (CuGGHLis). Kinetic characterization of inhibitor binding with fluorogenic assays showed that CuGGHLis binding was potent and non-selective (Ndom Ki =44.94±1.84nM; Cdom Ki =15.57±1.30nM). Catalytic inactivation was subsequently studied by monitoring remaining enzyme activity over time in the presence of redox co-reactants and CuGGHLis at IC20. In the presence of H2O2 and ascorbate, minimal diffuse radical oxidation of the Ndom was observed. Upon addition of CuGGHLis, however, the Ndom was rapidly and completely inactivated. Interestingly, the Cdom displayed greater diffuse radical oxidation and none catalysed by CuGGHLis. Insight into the mechanism was gained by characterizing residual activity through Michaelis Menten kinetics. This revealed that CuGGHLis catalyzed oxidation of Ndom residues involved in binding and hydrolysis of the non-domain selective substrate whereas Cdom residues were unaffected. No oxidative protein cleavage was detected by SDS-PAGE and silver staining, suggesting inactivation through side chain oxidation. Mass spectrometry will reveal the sites of metal-catalysed oxidation while X-ray crystallography will identify the CuGGHLis binding site. Time-course kinetic assays described the Cdom's lower thermal stability, higher dependence on detergent stabilization and greater diffuse radical oxidation. This is thought to be due to the unique hinging behaviour and higher carbohydrate content of the Ndom. Both domains were fully glycosylated with complex glycans, explicitly solvated and subjected to molecular dynamics simulations. While the Cdom active site is exposed to diffuse radicals, the Ndom is shielded by a glycan interaction network surrounding the molecule. This study provides further insight into the mechanism of N-selective irreversible inactivation by CuGGHLis. In future, these findings will aid the design of a new class of selective antifibrotic ACE inhibitors effective at sub-stoichiometric doses.

Speaker: Ms Lizelle Lubbe (University of Cape Town)

12:10 Lead optimization: calculating relative binding free energies with FEP+

The prediction of relative free energies of binding within a congeneric ligand series can be a potential key contribution to the process of ligand optimization. When accurate enough (approx. 1 log unit in the binding constant), free energy predictions can help to reduce the effort spent in compound synthesis and also the total time spent by reducing the number of iterations in the design process. Traditionally, fast empirical approaches like rigid-receptor docking dominated the field for decades due to their low computational cost, but have proven to be very limited in their predictive power in lead optimization and often required a large amount of experience to interpret since the success of the results can be highly target specific. Nowadays we are witnessing the beginning of a new era for relative binding free energy predictions due to the emergence of free energy perturbation (FEP) calculations. The general concept is quite old, and the first calculations on small systems were carried out more than twenty years ago. Thanks to a number of scientific and technological advances, we are now finally able to use FEP on a broader scope in real world applications. The talk will briefly outline the main features of Schrodinger’s FEP+ solution, which combines a number of recent advances in the field. FEP+ features high performance MD code that makes use of GPU computing, augmented by enhanced sampling schemes (‘replica-exchange solute tempering’), to achieve sufficient sampling to provide converged free energy estimates with modest hardware investment. Error analysis based on cycle-closure correction provides a measure of reliability for the calculations. On the force field side, the accurate all-atom force field OPLS3 is used, including a protocol to conveniently parameterize missing torsional parameters. Together with an automated setup procedure, this enables free energy calculations within real-world industrial projects. Beyond the technical overview, the presentation will outline a few interesting cases from literature and internal efforts where FEP+ has been used retrospectively and prospectively in order to give an idea about Schrödinger’s efforts in extending the domain of applicability of the technique.

Speaker: Dr Stephan Ehrlich (Schrodinger GmbH)

11:00 → 12:30 Cognitive Computing and Machine Learning: I

Convener: Prof. Fulufhelo Nelwamondo (CSIR)

11:00 Machine learning tools for visualising big data.

Unsupervised machine learning can be used to infer the hidden relationships inside of big data where there exists unknown structure and frameworks. Component based analysis seeks to reveal the correlation and variation within a dataset, processing and understanding these results can be challenging. 3D and 2D visualisation is used as a tool for expressing these n-th dimensional results in a simple and easily understood fashion. Unorganised streaming data separated into its principle components reveals anomalies and outliers which can be quickly detected to prevent data corruption.

Speaker: Prof. Amit Mishra (University of Cape Town)

11:30 Machine Learning in the Era of Data Intensive Astronomy

The construction of MeerKAT and emergence of the African VLBI network marks the beginning of the radio astronomy big data revolution in South Africa, and the first steps of the scientific and data pathway to the Square Kilometre Array (SKA). This journey to the SKA represents one of the most significant data challenges in scientific research of the coming decade. To rise to the scientific opportunity of this new data era requires the research community to develop new infrastructure, software systems and algorithm to process, analyse and mine these data for scientific knowledge. But the size of the data, the complexity of the signals contained therein, and the sophistication of the analytics needed, means we cannot rely on the old approaches for human interaction with the data. Machine learning techniques may be an important part of the solution. I will review the data-to-science landscape of the SKA, and touch on specific challenges where algorithmic approaches are falling short but machine learning may offer a way forward.

Speaker: Prof. Russ Taylor (Inter-University Institute for Data Intensive Astronomy)

12:00 Will AI transform your research and software?

Speaker: Mr Kiernan Mike (Microsoft)

11:00 → 12:30 Material Science: I

Convener: Dr Cornie van Sittert (North-West University)

11:00 MODELLING ADVANCED MATERIALS WITH HIGH PERFORMANCE COMPUTING

Modelling and simulation are now essential and increasingly important techniques in materials science. We aim to show the scope and challenges of the current field by reviewing recent work of the UK High End Computing Materials Chemistry Consortium. We will describe recent applications using both force-field and electronic structure techniques to key functional materials, including catalytic systems, materials for renewable energy technologies, electronic materials, ceramics and molecular materials. We will consider some of the future prospects and developments in the field.

Speaker: Prof. Richard Catlow (Department of Chemistry, University College London)

11:30 Tuning the electronic, magnetic and optical properties of 2D materials: Ab initio insights

The past decade has witnessed unprecedented advances in ab initio modelling of material properties. These advances have largely been attributed in part to algorithm refinements within codes implementing standard ab initio approaches such as density functional theory (DFT) and also to the advances made in high performance computing. It is thus now possible to model systems that were traditionally computationally expensive such as defects, extended surfaces and even nanoparticles. In this study the application of DFT to investigate the electronic, magnetic and optical properties of pristine and doped transition metal dichalcogenides (TMDCs), holey doped nitrogenated graphene g-C2N and graphene like GaN (g-GaN) is presented. Transition metal (TM) and lanthanide ion doping on the electronic, magnetic, optical properties of materials have also been taken into account. A more recent study on van der Waals hetero-structures/nanocomposites of graphene and graphene like AlN (g-AlN) for Li-ion battery (LIB) anode applications will also be presented in summary.

Speaker: Dr Cecil Ouma (North West University)

11:50 Computational study of anatase TiO2 as an anode material energy storage devices

The increasing demand for powering systems of portable electronic devices and zero-emission vehicles stimulates research towards high energy and high voltage systems is a challenge. Given these exciting properties, it becomes necessary not only to synthesize such solid-state and molecular systems but also to model their properties at an appropriate size and time scale. In this work we have predicted the structural parameters for TiO2 polymorphs and these parameters were found to be in agreement with an experimental data. We also calculated the band gap energies, predicted band structures and density of states for these polymorphs in an effort to validate the DFTB+ potentials. The geometry optimizations were performed using DFTB+ potentials that we derived. Moreover, these properties will determine which TiO2 polymorph can be used as an anode material in future storage devices.

Speaker: Mr Mulatedzi Gandamipfa (University of Limpopo)

12:10 Effect of pressure on structural, mechanical, dynamical and electronic properties of ReSe2 :A theoretical investigation

Effect of pressure on structural, electronic, mechanical, and Dynamical properties of ReSe 2 has been investigated using the first-principles calculations.These calculations are based on density functional theory and were performed using the generalized gradient approximation with the empirical van der Waals correction. The calculated lattice parameters ReSe2 under effect of pressure are consistent with the available experimental values upto 9.03 GPa, which motivated us to explore the stated prperties beyond the 9.03 Gpa limit. The lattice parameters and volume of ReSe2 decrease with the increasing pressure whereas the total enthalpy increases. The mechanical properties are predicted using the Voigt–Reuss–Hill approximation from which there is an inverse relationship between volume and pressure implying that as we subject ReSe2 to large pressure the material becomes somehow more compact due to reduced inter-atomic distance hence increase in magnitude of elastic coefficients as well as bond strength. The Dynamical instability was investigated by computing the phonon frequency to check for imaginary modes. ReSe2 is dynamically stable below 40 GPa. The band gap was estimated using the Perdew-Burke-Ernzerhof (PBE) and modified Becke–Johnson potential, which confirmed that as pressure increased, the band gap decreased in almost a linear manner.

Speaker: Mr Mahmoud Mahmoud (The University of the Witwatersrand)

11:00 → 12:30 SADC Meeting: I

SADC Cyberinfrastructure.

11:00 → 12:40 Storage and I/O: I

Understanding and Customizing Storage Systems:

General purpose storage systems aim to offer a least common denominator that works well enough for a variety of workloads. With inevitable failures and potential data loss at scale, it is important to understand what the chances are for data loss to minimize the probability. With this kind of knowledge, robust storage with interfaces customized to particular application requirements can be deployed. Through programmable storage systems interfaces, custom service features can be combined limiting overheads from unnecessary services. Using these tools, new kinds of storage services focused on data contents and quality of service requirements can be addressed. This session presents talks investigating understanding storage system component failures, programmable storage, and user-defined metadata services.

Organiser: Jay Lofstead (SANDIA)

Conveners: Dr Jay Lofstead (Sandia National Laboratories), Dr Peter Braam (-)

11:00 Scientific Modeling of Storage System Reliability

When designing large scale storage systems, failure is always a serious concern that demands constant attention. However, the ability for system designers to objectively evaluate their risk of data loss for a given storage system is minimal. Instead, they must enumerate possible failure modes, estimate their relative probability, identify possible mitigations, and decide whether the expense is worthwhile. This process relies on folk wisdom, rules of thumb, and anecdotal experience. For systems that grow larger and more complex year-by-year, this methodology is too imprecise to guarantee safety while ensuring efficiency. This talk will detail some of the progress in the SIMS^2 project, a collaboration between Sandia National Laboratories, University of Wisconsin-Madison, and Los Alamos National Laboratory charged with increasing the science and rigor behind evaluating system designs. It will cover some of the pitfalls of current methods of evaluating systems, methods for determining complex behavior of aggregated components, evaluation of different types of failure modes, and some interesting inflection points for different system designs.

Speaker: Dr Matthew Curry (Sandia National Laboratories)

11:25 Programmable Storage Systems

Storage systems need to support high-performance for special-purpose data processing applications that run on an evolving storage device technology landscape. This puts tremendous pressure on storage systems to support rapid change both in terms of their interfaces and their performance. But adapting storage systems can be difficult because unprincipled changes might jeopardize years of code-hardening and performance optimization efforts that were necessary for users to entrust their data to the storage system. We introduce the programmable storage approach, which exposes internal services and abstractions of the storage stack as building blocks for higher-level services. We also build a prototype to explore how existing abstractions of common storage system services can be leveraged to adapt to the needs of new data processing systems and the increasing variety of storage devices. We illustrate the advantages and challenges of this approach by composing existing internal abstractions into two new higher-level services: a file system metadata load balancer and a high-performance distributed shared-log. The evaluation demonstrates that our services inherit desirable qualities of the back-end storage system, including the ability to balance load, efficiently propagate service metadata, recover from failure, and navigate trade-offs between latency and throughput using leases.

Speaker: Prof. Carlos Maltzahn (University of California, Santa Cruz)

11:50 EMPRESS--Extensible Metadata PRovider for Extreme-scale Scientific Simulations

Significant challenges exist in the efficient retrieval of data from extreme-scale simulations. An important and evolving method of addressing these challenges is application-level metadata management. Historically, HDF5 and NetCDF have eased data retrieval by offering rudimentary attribute capabilities that provide basic metadata. ADIOS simplified data retrieval by utilizing metadata for each process' data. EMPRESS provides a simple example of the next step in this evolution by integrating per-process metadata with the storage system itself, making it more broadly useful than single file or application formats. Additionally, it allows for more robust and customizable metadata.

Speaker: Dr Jay Lofstead (Sandia National Laboratories)

12:15 HPC Storage – the two distinct challenges

Traditionally when architecting the storage for an HPC environment the focus is on performance of the file system. There is another challenge though, as one has to manage all the files that are used for input or created as output, typically over a long period of time. There is even value in this data that can be mined via modern analytics methods (aka Hadoop) or used for AI/ML. This session will outline why it is essential to use different solutions to address “the performance challenge” and “the persistent data challenge”. Biography: Wolfgang holds a PhD on computer aided engineering. In his more than 25 years in the IT industry he held various technical and strategic positions. He focuses on understanding the verticals’ specific processes, needs and applications, which he then links to the IT product’s features and capabilities in order to achieve the best results for the customers. In his current role as CTO for EMEA in Dell EMC’s Unstructured Storage Team he specializes mainly HPC and related verticals. Wolfgang also follows the immense transformations IT is currently undergoing; this especially includes the new capabilities data lakes, big data analytics and AI/ML offer.

Speaker: Dr Wolfgang Mertz (CTO HPC Dell EMC Isilon)

12:30 → 13:30 Lunch

13:30 → 15:00 Cognitive Computing and Machine Learning: II

Convener: Prof. Amit Mishra (University of Cape Town)

13:30 Computational Intelligence and Algorithmic Mechanism Design: an Application in Revenue Optimisation

Mechanism Design lies in the area of game theory, with aims to design games whose equilibria have desired objectives such as high efficiency or revenue optimisation. Algorithmic Mechanism Design focuses on Mechanism Design in algorithmically-complex scenarios, and it employs various analytics tools with considerations on computational constraints that exist in polynomial time. What makes this area more challenging is the trade-off between computational constrains and game-theoretic constrains. This talk will present this challenge, through a simple case study in the application of revenue optimisation in the area of auctions. A number of issues such as the computational hardness; convexity and a Vickrey–Clarke–Groves (VCG) mechanism, which is a generic truthful mechanism for achieving a socially-optimal solution, will be presented. The convergence of the challenges and the need for High Performance Computing environment will also be presented. The talk will end by discussing the current state of the field and how mechanism design is applied in computational settings, the challenges for further research.

Speaker: Prof. Fulufhelo Nelwamondo (CSIR)

14:00 Intel HPC and AI Technologies

In this presentation, attendees will hear about the latest Intel Technologies in AI and HPC. This will include Processors, Fabric, Accelerators, and some s/w as well. Also, I will touch base on some of Intel recent strategic acquisitions (e.g. Nervana, MobileEye, and Movidius). I will also include some benchmarks on different CPU's/applications that we tested.

Speaker: Mr Aljeshi Ahmed (Intel Corporation)

14:30 Accelerating Big Data Processing and Associated Deep Learning on Modern HPC Clusters and Clouds

The convergence of HPC, Big Data, and Deep Learning is becoming the next game-changing business opportunity. Apache Hadoop, Spark, gRPC/TensorFlow, and Memcached are becoming standard building blocks in handling Big Data oriented processing and mining. Modern HPC bare-metal systems and Cloud Computing platforms have been fueled by the advances in multi-/many-core architectures, RDMA-enabled networking, NVRAMs, and NVMe-SSDs during the last decade. This talk will provide an in-depth overview of the architecture of Hadoop, Spark, gRPC/TensorFlow, Memcached, and the associated Deep Learning frameworks. We will examine the challenges in re-designing networking and I/O components of these middleware with modern interconnects, protocols (such as InfiniBand, RoCE) and storage architectures. This talk will provide case studies of the new designs for several Hadoop/Spark/gRPC/TensorFlow/Memcached components and their associated benefits. Through these, we will also examine the interplay between high-performance interconnects, storage (HDD, NVM, and SSD), and multi-core platforms (e.g., Xeon x86, OpenPOWER) to achieve the best solutions for these components and applications on modern HPC clusters and clouds. We also present in-depth case-studies with modern Deep Learning tools (e.g., Caffe, TensorFlow, BigDL) running over RDMA-enabled Hadoop, Spark, and gRPC.

Speaker: Dr Xiaoyi Lu (The Ohio State University)

13:30 → 15:00 Earth Science: I

Convener: Dr Nicolette Chang (CSIR)

13:30 High Performance Computing at the South African Weather Service

The South African Weather Service (SAWS) is the national weather forecast and climate prediction agency and only atmospheric alerting authority in South Africa. The SAWS runs numerical models to simulate weather and climate over a variety of time scales, ranging from very short-range (0-12 hours) to multi-decadal timescales in what is termed a “seamless forecasting system” for the successful execution of its mandate. In 2013, the SAWS mandate was amended to also include air quality monitoring and modelling. The SAWS also runs marine and ocean models, as well as application models to help in decision making in sectors such as agriculture, water resource management, health, aviation, as well as energy. A critical enabling technology utilised by SAWS to perform the simulations described is a High Performance Computing (HPC) system and SAWS currently has a CRAY XC30 system. The current SAWS system is not enough for SAWS to conduct both its operational activities and research activities. Furthermore, the SAWS does not have a failover system to ensure business continuity in the event that the CRAY system fails to run. In order to address these issues, the SAWS and the Centre for High Performance have signed Memorundum of Agreement which will allow the SAWS to use the CHPC cluster as a fail-over system as well as for research purposes. Six research programmes that will address different research areas and timescales that the SAWS is working on have been agreed on, namely, 1) Very Short-Range Forecasting, 2) Short and Medium Range Forecasting, 3) Long-Range Forecasting, 4) Climate Change and Variability, 5) Air Quality Studies as well as 6) Applications Research. The talk will provide an overview on planned work, work in progress and work already done on the CHPC cluster in the different research programmes of the SAWS.

Speaker: Dr Mary-Jane Bopape (South African Weather Service)

14:00 South African Record Rainfalls

South African and world rainfall records are compared and new 5-minute, 10-minute, 15-minute and a possible yearly records for SA is reported. 42.4 mm, 70.8 and 76.0 mm were recorded by an Automatic Weather Station (AWS) at Levubu in Limpopo for 5, 10 and 15 minutes respectively from a thunderstorm 26 February 2016. A rainfall of 4299 mm was recorded for the year 2000 at Matiwa in Limpopo (Makhado district) from Tropical Cyclone Eline and other tropical systems that affected the area. Mapumulo Prison Automatic Rainfall Station (ARS) in eastern KwaZulu-Natal recorded 138.2 mm/hour 16:15 and 16:25 twice 27 January. Mapumulo's daily rainfall was 186.6 mm and the ARS lies halfway between Stanger and Kranskop. It is the SA record rainfall intensity for one hour pending the validation of the 218 mm in 40 minutes at Swartkoppies 22 km to south of Johannesburg 13 January 1951 recorded by an 8 inch standard rain gauge. The clock hourly rainfall record according to the Climate database is 125 mm 19:00 5 March 1997 at Bisho. Followed by 123 mm 16:00 at Mapumulo. A cloudburst occurs if rainfall rate is equal to or greater than 100 mm per hour.

Speaker: Mr Jan Vermeulen (South African Weather Service)

14:15 Modelling the variability of convection over East Rand South Africa

Speaker: Mr Sifiso Mbatha (South African Weather Service)

14:30 Computations on the CHPC: Evaluation of the COSMO model

The South African Weather Service (SAWS) is the mandated national meteorological service tasked with providing timely and accurate scientific data in the field of meteorology to the broader South African community, through a combination of both public and commercial services. It is, therefore, of great importance that the SAWS issues weather forecasts of high quality and value. SAWS accomplishes this task, partially, through the use of weather forecast models, with Unified model (UM) from the UK Metoffice serving as the main operational model. Weather forecast models perform differently across different parts of the globe. In order to ensure continuous delivery of high quality weather forecasts to SAWS stakeholders, model inter-comparisons are necessary. The main purpose of model inter-comparison is to examine, compare and analyse the outcome of different models in order to identify the most suitable model for SAWS. Previous studies have also shown that forecasts from multi-model systems are more skilful than those from one model. In addition to running the UM for operations, the SAWS is interested in studying the performance of other models over the country. Two models used in a number of countries have been chosen for this purpose and they are the Weather Research and Forecasting (WRF) and Consortium for Small-scale Modelling (COSMO) models. This study will focus on the COSMO model, which is a limited area model driven from Icosahedral Non-hydrostatic (ICON) global model. The ICON model runs with a grid spacing of 13km globally, and therefore COSMO cab be set up to run with a grid spacing that is smaller than 5km and therefore provide higher resolution simulations. The COSMO model is being run with a grid spacing of 2.8km and 50 vertical levels in Germany, which allows for accurate numerical prediction of near-surface weather conditions (e.g. clouds, fog, frontal precipitation) and simulation of severe weather events triggered by deep moist convection (supercell thunderstorms, intense mesoscale convective complexes, prefrontal squallline storms and heavy snowfall from wintertime mesocyclones). In this study, the COSMO model will be used to simulate selected case studies, and verified against the SAWS observations.

Speaker: Mrs patience tlangelani mulovhedzi (scientist)

14:45 Heatwaves in the future warmer climate of South Africa

Heatwaves in the future warmer climate of South Africa

Speaker: Mr Innocent Mbokodo (South African Weather Service)

13:30 → 15:00 Material Science: II

Convener: Prof. Eric van Steen (Catalysis Insititute, Department of Chemical Engineering, University of Cape Town)

13:30 DENSITY FUNCTIONAL STUDY OF ELECTRONIC, OPTICAL AND ADSORPTION PROPERTIES DYE ON ANATASE AND BROOKITE TIO2 SURFACES FOR APPLICATION IN DYE SENSITIZED SOLAR CELLS

Theoretical and computational studies of doped TiO2 polymorphous can contribute to a deeper understanding of dye sensitized solar cells. These solar cells represent a promising approach to a direct conversion of sunlight into electrical energy at low cost and with high efficiency. The light adsorption occurs in dye molecules adsorbed on a highly porous structure of TiO2 film [1-2]. The problem encountered with the TiO2 is its wide band gap which is about 3.4 eV and show photocatalytic activity under UV light irradiation that accounts for only a small portion of solar energy, in contrast to visible light which has a major part of solar energy. Harnessing and effectively utilizing sunlight is the most challenging subject for the extensive application of TiO2 as photon absorption [3]. Transition metal doping is one of the most effective approaches to extend the absorption edge of TiO2 to visible light region, which either inserts a new band into the original band gap, or modification of the conduction band (CB) or valence band (VB) improving the photocatalytic activity of TiO2 to some degree [3-5]. In the current study, electronic density of states and optical properties of doped and undoped, Anatase and Brookite surfaces were performed using the first-principles calculations based on Density Functional Theory (DFT) using a plane-wave pseudopotential method. The generalized gradient approximation (GGA) was employed in the scheme of Perdew-Burke-Ernzerhof (PBE) to describe the exchange-correlation functional. All calculations were carried out with CASTEP (Cambridge Sequential Total Energy Package) code in Materials Studio of Accelrys Inc [6]. The results confirm that the mixing of the dopants induced states with the original Ti 3d and O 2p valence band and conduction band attributes to the band gap, hence the shifting of the absorption edge of TiO2 from UV to visible spectrum. The light harvesting efficiency of the dye molecules were calculated and compared to the experimental values. Keywords: Dye sensitized solar cells, Density Functional theory, Bandgap, Optical and Electronic Properties. References 1. Chou T.P., Zhang Q., Cao G., (2007), Effects of dye loading conditions on the energy conversion efficiency of ZnO and TiO2 dye-sensitized solar cells. Journal of Physical Chemistry C, 111, 18804-18811. 2. Yuasa T., Kawakami R., Sato Y., Mori Y., Adachi, M., Yoshikado, S., (2012), Dye adsorption for dye-sensitized solar cell, Solar Energy Materials and Solar Cells, 102, 2-7. 3. Zallen R, and Moret M.P. The optical absorption edge of brookite TiO2. Solid State Commun, 2006, 137 154-157. 4. Yang K, Dai Y, Huang B, Han S: Theoretical study of N-doped TiO2 rutile crystals. J Phys Chem B 2006, 110:24011–24014. 5. Grätzel M, and O’Regan B. “A Low-Cost, High-Efficiency Solar Cell Based on dye sensitized Colloidal TiO2 Films” Nature, 353, 737-740, 1991. 6. Clark S.J, Segall M.D, Pickard C.J, Hasnip P.J, Probert M.I.J, Refson K and Payne M.C 2005 Z Kristallogr 220 567–570.

Speaker: Dr Eric Maluta (University of Venda)

14:00 Charge transport, interfacial interactions and synergistic mechanism in BiNbO4/MWO4 (M = Zn and Cd) heterostructures for hydrogen production: insights from a DFT+U study

Speaker: Mr Francis Opoku (Department of Applied Chemistry, University of Johannesburg)

14:20 STRUCTURE PROPERTY RELATIONSHIPS IN ORGANIC SOLAR CELLS MATERIALS

Speaker: Mr Sandile Mamba (UKZN)

14:40 Beyond Li-ion: Computational Modelling Studies on Stability of Li-S-Se System

Recent rechargeable batteries are mainly based on conventional lithium intercalation chemistry, using lithium transition metal oxides as cathode material with typical capacities of 120-160 mA.h/g. The low energy density and/ or high cost of these cathode materials have limited their large scale production and application in Li ion batteries. Exploration of new cathode materials is consequently necessary to realise more efficient energy storage systems. Lithium sulphur cells have a promise of providing 2-5 times the energy density of Li-ion cells, however, they suffer poor cycling performance. Improvements that are effected by using Li/SeSx system in different electrolytes have been reported. In the current study we employ computational modelling methods to explore stability, structural and electronic properties of discharge products formed in the Li/SeSx battery, which has potential to offer higher theoretical specific energy and remedies the challenges that Li-S battery encounters. First principle methods were used to calculate thermodynamic properties of Li2S and Li2Se, which agreed with available experimental results. A cluster expansion technique generated new stable phases of Li/SSex system and Monte Carlo simulations determined concentration and temperature ranges in which the systems mix. Interatomic Born Meyer potential models for Li2S and Li2Se were derived and validated and used to explore high temperature structural and transport properties of mixed systems.

Speaker: Mr Cliffton Masedi (CSIR/UL)

13:30 → 15:00 SADC Meeting: II

SADC Cyberinfrastructure.

15:00 → 15:30 Break

Refreshment break

15:00 → 15:30 Chemistry and Materials SIG

Special Interest Group Meeting for Computational Chemistry and Materials Science

Conveners: Dr Cornie van Sittert (North-West University), Prof. Penny Govender (Department of Applied Chemistry, University of Johannesburg)

15:30 → 16:30 Plenary: Vendor Crossfire

Convener: Mr Michael Feldman (TOP500.org)

15:30 2017 HPC Landscape and Technology Trends

As the exascale era approaches, the landscape of high performance computing is changing rapidly. The advent of AI and machine learning has emerged as a major influence in HPC system and component design, and has encouraged users to rethink their traditional workflows. TOP500 News Editor Michael Feldman will recap some of the most important developments in this area that took place over the last year and discuss the broader ramifications for the HPC community.

Speaker: Mr Michael Feldman (TOP500.org)

18:00 → 18:45 Plenary: Diamond Sponsor

Convener: Dr Happy Sithole (CHPC)

18:00 Intel Plenary TBA

Speaker: Dr Bill Magro (Intel)

18:45 → 21:00 Dinner: Braai

Tuesday, 5 December

08:00 → 09:00 Registration

09:00 → 10:30 Plenary

Convener: Mr Leon Staphorst (Director: SANReN)

09:00 The Demand for High Performance Computing in Kenya: Special Interest Groups as vehicles for promoting research collaboration

Kenya Education Network (KENET) is the National Research and Education of Kenya responsible for building and operating the research infrastructures for the academic community in Kenya. The research infrastructures include a high speed national network with low- congestion international links to global research & education networks including TENET / SANREN in South Africa. This talk will describe how KENET has created Special Interest Groups (SIGs) in order to discover and promote research collaboration in three main priority areas, namely, engineering, computational modelling and materials science (CMMS), and educational technology. The talk shall describe the innovative methods used to discover research champions in different areas and the joint activities of one of the Special SIG on CMMS in the period 2016-2017. Kenya plans to build the capacity of Kenyan researchers in the use of HPC for researchers with the support of CHPC and other grid infrastructures operators such as the European Grid Infrastructure (EGI). The talk shall describe how KENET has setup a low- cost Certification Authority (CA) that has been accredited by EUGRIDPMA. The certificates issued by the CA would allow Kenyan researchers to access EGI to supplement their computing needs. The talk concludes that any African NREN can facilitate access to HPC and Grid Computing facilities in any part of the world and therefore to promote research collaboration and productivity.

Speaker: Prof. Meoli Kashorda (KENET)

09:45 Dramatic Changes for HPC Storage and IO

Speaker: Dr Peter Braam (-)

10:30 → 11:00 Break

10:30 → 11:00 Poster

Students will be available to answer questions about their posters.

Convener: Dr Andrew Gill (CHPC)

11:00 → 12:30 Chemistry: II Chemistry & Health

Convener: Prof. Gert Kruger (UKZN)

11:00 Molecular modelling of ligands, receptors and enzymes for in silico drug discovery

Over the past two decades, computational or in silico methods have been increasingly applied to the process of drug development and testing. These methods include the use of quantitative structure-activity relationships, database searching, pharmacophores, homology models and other molecular modeling approaches. With the advent of cheminformatics techniques in drug design, molecules with promising efficacy can be developed in a comparatively short time span. Public chemical databases can be screened and the compounds predicted for behaviour as inhibitors or activators. The highly sophisticated pharmacoinformatics tools available can assist in the generation of even subatomic descriptors that provide high predictivity for activity.

Speaker: Prof. Tahir Pillay (University of Pretoria)

11:30 Using docking and molecular dynamics to design potent and selective anticancer, anti-osteoclastogenic and anti-malarial compounds

Docking and molecular dynanmics simulations using Autodock Vina and Gromacs were used to identify selective inhibitors of cancer-, osteoclastogenic- and malaria-associated proteins. Selective benzotriazepine inhibitors of bromodomain 4 were identified and synthesized. The compounds show micromolar growth inhibition of several cancer cell lines, as well as potent inhibition of osteoclastogenesis without cytotoxicity against osteoclast progenators. Potent inhibitors of malaria cytochrome bc1 proteins were identified and synthesized. The most potent compound inhibited in vitro asexual malaria parasite growth to 50% of the control at a concentration of 64nM. The compound showed no activity against human HepG2 cells at 5uM. The present study will discuss how freely available docking and molecular dynamics tools are able to help researchers narrow down new, potent and selective pharmaceutically relevant compounds.

Speaker: Dr Andre Stander (University of Pretoria)

11:50 The identification of highly potent peptide scaffolds for the treatment of hormone resistant prostate cancer

The risk of prostate cancer in men has grown significantly over the last 5 years while the chemotherapeutic market for prostate cancer treatment is a $5 billion dollar global industry. The growth of hormone sensitive cases of cancer treatment failure has motivated the rise of alternative approaches to its chemotherapeutic treatment. One such approach involves targeting the microtubule-associated protein/microtubule affinity-regulating kinase 4 (MARK4) protein. MARK4 is a kinase involved in signal transduction and is a possible chemotherapeutic target for prostate cancer treatment. Peptides possess a large chemical space that can rapidly identify potent scaffolds for drug design strategies. We populated a virtual library of over 50,000 unique, lead-like synthesisable peptides suited to drugging the MARK4 catalytic ATP recognition site using DerivatizeME an in-house development virtual library builder. High-throughput virtual screening of the virtual library identified 100 hits based on binding probabilities expressed as molecular docking binding energies. Rapid simulation based screening of these hits identified 10 highly potent peptide scaffolds for chemical synthesis in order to identify peptide inspired drug precursors. Further interrogation of these hits for the treatment of hormone resistant prostate cancer through perturbing signal transduction via targeting the MARK4 protein will enable us to be better positioned to win the war against prostate cancer.

Speaker: Mr Lester Sigauke (Rhodes University)

12:10 The enzymatic mechanism of Human Immunodeficiency Virus type 1 (HIV-1) protease: are we there yet?

**Introduction** The catalytic mechanism of the Human Immunodeficiency Virus type 1 (HIV-1) protease (PR) is one of the most studied aspartate protease representative. Both experimental and theoretical techniques have been harnessed to provide profound understanding on a number of possible reaction pathways for the catalysis of HIV-1 PR on its natural substrate/ligand. Most of these studies have investigated the stepwise general acid/base mechanism with little attention on a synchronous model in which the proteolytic reaction could occur as a one-step concerted process. Jaskólski et al. first put this proposal forward in 1991 in which the hydrolytic reaction is viewed as a one-step process; the nucleophile (water molecule) and electrophile (an acidic proton) attack the scissile bond in a concerted manner. **Aim and objective** Herein, the one-step concerted catalytic mechanism of HIV-1 PR on its natural substrate and a fluoride derivative was studied using density functional theoretical (DFT) method. **Method** The reaction was modeled to proceed through the formation of a six-membered ring transition state structure, which was facilitated by a pre-reaction enzyme-substrate complex at B3LYP/6-31+G(d) level of theory using Gaussian 09 program suite. The applied in silico model allows the elucidation activation parameters, kinetics, solvent contributions and quantum chemical properties for this system. **Result** Theoretically determined activation free energy of 19 kcal mol-1 obtained was very close to approximately 18 kcal mol-1 reported from experiment. The fluorinated peptide substrate has an activation free energy of 12 kcal mol-1 which is 7 kcal mol-1 lower than natural substrate. **Implications of result** This investigation could potentially serve as a basis towards understanding the enzymatic mechanism of homodimeric enzymes and could also guide future design of better HIV-1 PR inhibitors through fluorinating the scissile nitrogen of the natural substrates; an ongoing perspective from our research group.

Speaker: Ms Monsurat Motunrayo Lawal (UKZN)

11:00 → 12:40 Earth Science: II

Convener: Prof. Marius Potgieter (North-West University)

11:00 Ocean-biogeochemical model sensitivity in the Southern Ocean towards the development of the earth system model, CSIR-VR-ESM

Modelling the earth system is a balance between available computing power and model complexity, particularly as an earth system model comprises of multiple individual, interacting numerical models each representing different system components such as the atmosphere, land and ocean. In addition, these models are run for very long periods to simulate hundreds of years of earth/climate evolution. For the development of the CSIR Variable Resolution Earth System Model (VR-ESM), we discuss the compromises of computation versus complexity made in configuring the ocean-biogeochemistry component. The Southern Ocean is a key region for global carbon exchange: both physical and biological mechanisms drive carbon from the atmosphere to the surface ocean and then to the deep ocean and conversely. Our research shows that the spatial scales of the surface ocean dynamics are important in driving these exchanges, these range from the mesoscale O(10-100km) to the submesoscale O(1km). However, the ocean components of earth system models (and thus the associated biogeochemical model) were generally run at coarse resolutions of 100km or more, and have recently incorporated the mesoscale through model resolution. Submesoscale-resolving tends to be reserved for the local to regional ocean domain. Leveraging the representation of the necessary mesoscale and submesoscale ocean features over a global ocean with the computational power of the CHPC is a key decision in the configuring of VR-ESM. For the ocean model, this can be done by the choice of model resolution and/or the use of subgrid-scale parameterizations. We have used the ocean-ice-biogeochemistry modeling platform NEMO to perform sensitivity tests using varying model resolution and subgrid-scale parameterizations in order to best represent the necessary ocean-biogeochemical dynamics in the Southern Ocean. This leads to optimizing the ocean model configuration in VR-ESM with consideration to cpus and wall time.

Speaker: Dr Nicolette Chang (CSIR)

11:30 Prospects for improving coastal modelling in the Benguela region

Speaker: Dr Thulwaneng Mashifane (South African Environmental Observation Network)

11:50 Mechanisms of the Sea-Air CO2 Flux Seasonal Cycle biases in CMIP5 Earth Systems Models in the Southern Ocean

The Southern Ocean forms a key component of the global carbon cycle. Recent studies, however, show that CMIP5 Earth System Models (ESM) disagree on the representation of the seasonal cycle of the CO2 flux (FCO2) and compare poorly to observations in the Southern Ocean. This model-observations bias has important implications on the ability of ESMs to predict century scale CO2 sink and related climate feedbacks. In this study, we used a specialized diagnostic analysis on 10 CMIP5 models in the Southern Ocean to discriminate the role of the major drivers, namely the temperature control and the concentration of dissolved inorganic carbon (DIC). Our analysis shows that the FCO2 biases in CMIP5 models cluster in two major groups . Group A models (MPI-ESM-MR, NorESM2 and HadGEM-ES) are characterized by exaggerated primary production such that biologically driven DIC changes mainly regulate the seasonal cycle of FCO2. Group-B (CMCC-CESM, GFDL-ESM2M, IPSL-CM5A-MR, MRI-ESM, CanESM2, CNRS-CERFACS) overestimates the role of temperature and thus the change in CO2 solubility becomes a dominant driver of FCO2 variability. While CMIP5 models mostly show a singular dominant influence of these two extremes, observations show a modest influence of both, with a dominance of DIC regulation. We found that CMIP5 models overestimate cooling and warming rates during autumn and spring with respect to observations. Because of this, the role of solubility is overestimated, particularly during these seasons (autumn and spring) in group B models, to the extent of contradicting the biological CO2 uptake during spring. Group A does not show this solubility driven bias due to the overestimation of DIC draw down. This finding strongly implies that the inability of the CMIP5 ESMs to resolve CO2 biological uptake during spring might be crucially related to the sensitivity of the pCO2 to temperature in addition to underestimated biological CO2 uptake.

Speaker: Mr Precious Mongwe (CSIR NRE)

12:05 Modelling the low frequency variability of the Indian Ocean.

This study investigates the low-frequency variability of the Indian Ocean based on Coupled Model Intercomparison Project (CMIP5) model outputs, observed and reanalysis datasets, and CSIR ocean models. While the existence of semi-annual and annual signals is well known the presence of decadal to multi-decadal signals is less understood. In the South African context, the variability of the Indian Ocean associated with large-scale, low-frequency modes of variability, has implications for the predictability of the regional climate driven by ocean dynamics. Thus far, low pass filtered Ocean Temperature, Sea Surface height, and zonal and meridional components of the ocean currents are subjected to spectral analysis for identification of statistically significant periodicities at the low-frequency timescale. This study also aims to report the relationship between low-frequency variability in the Indonesian Throughflow (ITF) outflow region and the southern Indian Ocean (SIO) in the decadal to multi-decadal timescale.

Speaker: Mr Mthetho Sovara (CHPC/UCT/CSIR)

12:20 Understanding rainfall distribution over the Drakensberg Mountain Range under weak synoptic conditions.

Understanding rainfall variability under weak synoptic forcing is crucial for improving weather forecast and climate projections with Numerical Weather and climate models. Rainfall variability under strong synoptic-scale forcing is dictated by large scale features whereas under weak synoptic-scale forcing, it is dictated by local-scale features. However, most of these local scale features are not well resolved by low resolution numerical weather and climate models which are used for operational weather forecast and climate projections. Consequently, the models may underestimate or overestimate the intensity of climate extremes. The analysis of the characteristics of wet weak synoptic days from observation and reanalysis data over the Drakensberg Mountain Range (DMR) revealed that although the reanalysis data captures the observed characteristics of wet weak synoptics days, it underestimates the frequency and intensity because of its coarse horizontal resolution. Therefore, using Self-organizing maps (SOM) and high resolution simulations with the Weather Research and Forecasting (WRF) model, we investigated the characteristics of rainfall under weak synoptic scale forcing over the DMR. The results of the SOM classification of Wet Weak Synoptic Days rainfall revealed four main rainfall patterns under weak synoptic conditions over the DMR. The simulation of these rainfall patterns with the WRF model revealed that the WRF model greatly improves reanalysis data and brings it closer to observation and that a WRF model horizontal resolution of at least 18km is required to capture most of the local scale processes influencing rainfall formation under weak synoptic conditions. Hence, higher resolutions (less than 18km) simulations may not be necessary which implies a lower computational cost for simulations.

Speaker: Mr Ridick Roland TAKONG (University of Cape Town)

11:00 → 12:30 HPC: I: Techniques

Convener: Dr Davini Maurizio (University of Pisa)

11:00 Accelerating the 3-D FFT using a heterogeneous FPGA architecture

Future Exascale architectures will likely make extensive use of computing accelerators such as Field Programmable Gate Arrays (FPGAs) given that these accelerators are very power efficient. Oftentimes, these FPGAs are located at the network interface card (NIC) and switch level in order to accelerate network operations, incorporate contention avoiding routing schemes, and perform computations directly on the NIC and bypass the arithmetic logic unit (ALU) of the CPU. This talk explores just such a heterogeneous FPGA architecture in the context of two kernels that are driving applications in leadership machines: the 3- D Fast Fourier Transform (3-D FFT) and Asynchronous Multi-Tasking (AMT). The machine explored here is a DataVortex system which consists of conventional processors but with programmable logic incorporated in the memory architecture. The programmable logic controls the network and is incorporated both in the network interface cards and the network switches and implements a contention avoiding network routing. Both the 3-D FFT and AMT kernels show compelling performance for deployment to FFT driven applications in both molecular dynamics and density functional theory.

Speaker: Matthew Anderson (Indiana University)

11:25 Unconventional use of Intel XeonPHI Knights Landing

In this talk we will look at Knights Landing capabilities and its use for computations that are not strictly accelerator-oriented. The platform supports heterogeneous computations given the flexibility of their cores, allowing vector code to execute side by side along with more enterprise-oriented codes.

Speaker: Prof. Antonio Cisternino (University of Pisa)

11:50 Understanding and Controlling Processor Affinity

Speaker: Dr Milfeld Kent (TACC)

12:15 CODE-RADE offloading software building from sites.

CODE-RADE is a platform for user-driven, continuous integration and delivery of research applications in a distributed environment. Starting with 6 hypotheses describing the problem at hand, we put forward technical and social solutions to these. Combining widely-used and thoroughly-tested tools, we show how it is possible to manage the dependencies and configurations of a wide range of scientific applications, in an almost fully-automated way, via constant integration tools and delivery into cvmfs. Due to the complexity and number both of scientific packages as well as computing platforms, delivering these applications to end users has always been a significant challenge through the grid era, and remains so in the cloud era. The CODE-RADE platform is a means for developing trust between public computing and data infrastructures on the one hand and various developer and scientific communities on the other hand. Predefined integration tests are specified for any new application, allowing the system to be user-driven. This greatly accelerates time-to-production for scientific applications, while reducing the workload for administrators of HPC, grid and cloud installations together with the people maintaining the software. Specific examples will be given for the HPC facility in Cape Town and the distributed grid resources within South Africa. We will give some insight into how this platform could be extended to address issues of reproducibility and collaboration in scientific research in Africa. Finally, the mechanism of giving credit for work done which is citable will be explained.

Speaker: Mr Sean Murray (CHPC, CSIR)

11:00 → 12:30 Material Science: III

Convener: Dr George Amolo (Technical University of Kenya, Nairobi)

11:00 Computational modeling as a complementary and predictive tool in materials science research

With the recent scenario of affordable computers with large RAM and storage capabilities as well as the availability of high performance computing facilities, research in materials science using ab initio techniques has rapidly grown. The research focus has now gradually shifted from fundamental science investigations to properties of materials that have applications in environmental preservation, search for as well as adoption of alternative fuels and green energy at large, health, among others, which have an expected impact on the quality of life. In the last few years our group has worked on hard materials for cutting and shaping in industry as well as materials for energy conversion and materials for electronic applications. In particular, we have investigated hardness in Transition Metals Carbides and Nitrides (TMCNs) and extended to other indicators such as shear modulus, rigidity and brittleness; metal electrode contacts on diamond surfaces for diamond electronics; metal hydride materials for applications in hydrogen fuel storage and the electronic structure of the interface of dye/TiO2 in dye sensitized solar cells (DSSC), among others. Highlights of the finding of these cases and recent advances are provided. These examples, among others, are provided to show how computational modeling richly supports experimental work and yet can also be used to provide useful information that may not directly be accessible to experiments.

Speaker: Dr George Amolo (Technical University of Kenya)

11:20 Spin-transfer torque field switching of magnetization in vertically-stacked multilayers

Theoretical and computational analysis of the magnetotransport properties and spin-transfer torque field-induced switching of magnetization density in vertically-stacked multilayers is presented. Atom-resolved magnetic moments and spin-transfer torques are computed at finite bias within linear response approximation to the spin-density reformulation of the van der Waals density functional theory. Dynamical spin excitations are computed as a function of a spin-transfer torque induced magnetic field along the magnetic easy axis, and the corresponding spin polarization perpendicular to the easy axis is obtained. A giant anisotropic magnetoresistance of 3200% is obtained in the nonmagnetic-metal-capped Fe/hBN/graphene/hBN/Pt multilayer architecture. The magnitude of the spin-transfer torque is found to increase as the tunneling spin current increases, and this activates the magnetization switching process due to increased charge accumulation. This mechanism causes substantial spin backflow, which manifests as rapid undulations in the bias-dependent tunneling spin currents. The implication of these findings on the design of nanoscale spintronic devices with spin-transfer torque tunable magnetization density is discussed. Insights from this study are expected to enhance the prospects for developing and integrating artificially assembled van der Waals multilayer heterostructures as the preferred material platform for efficient engineering of spin switches for spintronic applications.

Speaker: Dr Aniekan Ukpong (UKZN)

11:40 Phase diagram for the adsorption of oxygen and water on Pt(100) and Pt(111)

Speaker: Mr Pierre Cilliers (Catalysis Insititute, Department of Chemical Engineering, University of Cape Town)

12:00 DMol3/COSMO-RS Prediction of Aqueous Solubility and Reactivity of Selected Azo Dyes: Effect of Global Orbital Cut-off and COSMO Segment Variation

Speaker: Mr Olaide Wahab (Department of Applied Chemistry, University of Johannesburg, P. O. Box 17011, Doornfontein Campus, 2028, Johannesburg, South Africa)

11:00 → 12:30 SA NREN: I

Convener: Dr Renier van Heerden (SANReN, CSIR)

11:00 Expanding the borders of Nelson Mandela University – Strategic partnerships to enable high speed university connectivity in Nelson Mandela Bay

The university landscape in South Africa has significantly changed since the latter part of 2015. Never before has issues of sustainability been more prevalent and universities are, more than ever, exploring strategic partnerships in order to sustainably provide connected student spaces in support of its academic agenda. Through creative and opportune partnerships with the Nelson Mandela Bay Metropole (NMBM) as well as Vodacom and MTN, the Nelson Mandela University was not only able to connect its satellite campuses with high-speed fibre, but also provide university connectivity from eleven public libraries across the metropole and the Nelson Mandela Bay World Cup Stadium. The new Wi-Fi hotspots within the NMBM libraries were set-up to be extensions of the Nelson Mandela University network and provides students with Wi-Fi services comparable to what they would receive through the campus computer labs. The Nelson Mandela Bay Stadium was connected to the university network within the span of one week and this connectivity, in conjunction with connectivity in public libraries, proved critical in salvaging the university’s 2016 academic year when its main campuses were closed due to student unrest.

Speaker: Mr Creswell Du Preez (Nelson Mandela University)

11:30 TVET Connectivity Project

SABEN is a non-profit focused on ending bandwidth poverty among Sooth Africa’s schools and colleges. It collaborates with the South African National Research and Education Network (SANReN) and procures services from the Tertiary Education and Research Network of South Africa (TENET) to provide high capacity bandwidth for schools networks and colleges. SABEN also offers other over-the-network services such as hosting, VOIP and video conferencing services.

Speaker: Mr Arno Hart (SABEN)

12:00 An Assessment of the Maturity of the SA NREN

The South African National Research and Education Network (SA NEN) is a collaborative effort by the Tertiary Education and Research Network of Of South Africa (TENET) and the South African National Research Network (SANReN) than has panned more than a decade. While the SA NREN initially ha the primary goal to provide the South African higher education and research communities with unconstrained bandwidth at a fraction of the costs of the commercial market, it has now evolved to also provide a wide variety of advanced services and participates in various international initiatives. As such, it has evolved from a mere connectivity provider to a an extremely advanced NREN that is now regarded as one of the world's leading Research and Education (R&E) networks, as is evident from it's membership in the Global R&E Chief Executive Office (CEO) Forum. This presentation will consider the progress that the SA NREN has made in recent times in providing connectivity and other advanced services to its beneficiairies, and measure this against the NREN maturity model developed by TENET to gauge the successful development of NRENs throughout Africa.

Speaker: Mr Leon Staphorst (Director: SANReN)

11:00 → 12:30 SADC Meeting: III

SADC Cyberinfrastructure.

12:30 → 13:30 Lunch

13:30 → 15:00 Chemistry: III

Convener: Prof. Ignacy Cukrowski (University of Pretoria, Department of Chemistry)

13:30 Computational chemistry methods in understanding the properties of porous material

Intermolecular interactions play a fundamentally important role in the properties of solid materials. For instance, molecules ("guests") are taken up into porous materials ("hosts") as a result of the interactions between these species, while the manner in which they interact has an influence on the sorption ability of the porous material. Several examples from our work will be used to show that calculations performed using the CHPC's computational facility allow us to explain the role that intermolecular interactions play in the unusual sorption properties of various porous compounds. For instance, the interactions between carbon dioxide and a host porous metal-organic framework yield anomalous sorption isotherms.[1] Similarly, intermolecular interactions are responsible for the change in colour of a crystalline porous compound during sorption of certain solvents along an hourglass pattern. Calculations show that the origin of this effect is that the channels in the porous framework are anisotropic, allowing sorption only from particular faces.[2] Calculations also enabled us to understand the changing intermolecular interactions involved in two other porous compounds that could allow the frameworks to undergo large structural changes during sorption and desorption, while still maintaining the integrity of the crystalline nature of the compound. [3] ---------- [1] Bezuidenhout, C. X.; Smith, V. J.; Bhatt, P. M.; Esterhuysen, C.; Barbour, L. J. Angew. Chem. Int. Ed. 2015, 54, 2079–2083. [2] Bezuidenhout, C. X.; Esterhuysen, C.; Barbour, L. J. Chem. Commun., 2017, 53, 5618–5621. [3] Bezuidenhout, C. X.; Smith, V. J.; Esterhuysen, C.; Barbour, L. J. J. Am. Chem. Soc. 2017, 139, 5923–5929.

Speaker: Prof. Catharine Esterhuysen (Department of Chemistry and Poymer Science, University of Stellenbosch)

14:00 Can we reach the heaven of chemical accuracy by climbing Jacob’s Ladder? A DFT validation study for Non-heme Iron complexes

Speaker: Dr Adedapo Adeyinka (Chemical Resource Beneficiation Focus Area, Northwest University Potchefstroom Campus)

14:20 Molecular modelling of lanthanide(III) chloride complexes with the nitrogen ligand DPA: AIM/NCI analysis and ligand conformational search

**Introduction** Geometry optimizations were performed on the crystal structure asymmetric units as determined for the novel lanthanide coordination compounds: LaCl$_{\text{3}}$(DPA)$_{\text{2}}$, µ-Cl$_{\text{2}}$-[PrCl$_{\text{2}}$(H$_{\text{2}}$O)(DPA)]$_{\text{2}}$, [LnCl$_{\text{2}}$(DPA)$_{\text{2}}$]Cl (Ln = Nd,Dy & Y). Lanthanide complexes are known to be stereochemically labile, with variable coordination numbers ranging from 2-12. This is due to the localization of f-electrons near the core region of the metal nucleus and the associated ligand-to-metal ionic bonding. The conformation of the ligands would therefore be highly dependent on their respective environment, i.e. on the strength of both intra- and intermolecular interactions. [1] This study will focus mainly on identifying and characterizing intramolecular interactions that contribute to the overall stability of the geometry of the complexes, in the gas-phase (*in vacuo*). Hydrogen bonding being one of the most important examples of stabilizing interactions. **Theoretical methods** All geometry optimisations were carried out with the *Gaussian 09* [2] software package. The geometry optimisation calculations, of the lanthanide complexes, were performed using density functional theory (DFT) at the M06-D3/aug-cc-pVTZ/RECP level. For each lanthanide metal center the ‘core’ electrons were represented by a small-core (28 electron) Stuttgart-Dresden (SDD) quasi-relativistic effective core potential (ECP), while the remaining ‘valence’ electrons (incl. f-electrons) were explicitly modelled using the energy-optimized (14s13p10d8f6g)/ [6s6p5d4f3g] atomic natural orbital (ANO) Gaussian valence basis sets of Cao and Dolg .[3] Atoms-in-molecules (AIM) electron density topological analysis [4], shown in Fig. 1, and non-covalent interaction (NCI) analysis [5], shown in **Fig. 2**, were performed on the energy minimized structures, using the software package Multiwfn [6]. Initial optimizations on DPA were performed using the M06-D3/6-311++G(d,p) level, followed by the larger aug-cc-pVTZ basis set. **Ligand conformational search** The ligand geometry optimizations were successfully performed on 621 initial geometries, which were generated by identifying all unique permutations of four principle dihedral angles, in 60° increments, found in the N-donor ligand *bis*(pyridin-2-ylmethyl)amine (*DPA*). The total number of input geometries were reduced by accounting for the inherent symmetry present in the *DPA* molecule, in order to prevent the redundant optimisation of equivalent initial geometries. **AIM/ NCI analysis** One of the principle components of the atoms-in-molecules (AIM) approach involves locating stationary points in space where the curvature of electron density are characterised according to the following four categories [10], see **Fig.1**: - Nuclear critical points [NCPs, (3, -3)] - Bond critical points [BCPs, (3, -1)] - Ring critical points [RCPs, (3, +1)] - Cage critical points [CCPs, (3, +3)] Non-covalent interaction (NCI) analysis supplements the aforementioned approach, particularly where weak interactions may exist in the absence of BCPs. [5] This approach identifies regions in space where attractive interactions (sign(λ$_{\text{2}}$)ρ<0) or repulsive interactions (sign(λ$_{\text{2}}$)ρ>0) dominate, see **Fig. 2**. **Acknowledgements** The author would like to thank Dr. Eric C. Hostena for solving the crystal structures of the aforementioned inorganic complexes, via single-crystal XRD, as well as Dr. Abubak’r Abrahams for his support. The author would like to acknowledge the Centre for High Performance Computing (CHPC) of CSIR, in South Africa, for enabling access to usage of computational resources to conduct this research.

Speaker: Mr Cameron Matthews (Nelson Mandela University)

14:40 A DFT study of the ring opening mechanism of tetraethyl-2-aryl-1,2-epoxygembisphosphonates

A DFT investigation of the ring opening mechanism of tetraethyl 2-aryl-1,2-epoxygembisphosphonates on the C-O bonds of the oxirane ring using ammonia as the nucleophile were carried out. Thermodynamic data obtained in both the gas phase and solvent simulations showed that the ring opening is favoured at the less hindered carbon (C2) of the oxirane ring. Varia-tion of the substituent on C2 to lower the activation energy revealed electron-donating methoxybenzene to be the best sub-stituent in this study.

Speaker: Dr Thishana Singh (DUT)

13:30 → 15:00 Cognitive Computing and Machine Learning: III

Convener: Mr Kiernan Mike (Microsoft)

13:30 Deep Learning: Research Directions and Upcoming Challenges

The focus of this talk is an analysis of the upcoming trends in Deep Learning research, in particular as regards the most promising application fields which are likely to benefit from the AI revolution. We will take a deep dive into Deep Learning applications to Life Science and the Internet of Things, highlighting the challenges and advantages related to working on data with a structured representation. We will show examples ranging from sensor time series to labelled trees representing subclonal cancer populations, all the way to more complex graph structures capturing the articulated interactions between brain areas during cognitive processes. Particular focus will be given to the challenges and opportunities for the HPC and application developers communities.

Speaker: Dr Davide Bacciu (University of Pisa)

14:00 Applications of machine learning: Developing natural language processing technology for South Africa's official languages

Recent developments in the field of machine learning (ML) have been influenced by advances in computational technology as well as the availability of large volumes of data. One field in which progress has been accelerated substantially is Natural Language Processing (NLP). NLP is a broad research field including topics such as natural language understanding and natural language generation. The aim of research in this field is to enable natural language interaction between humans and machines. This talk will present some of the ML techniques involved in NLP and will provide a brief overview of how things have changed as a result of Big Data and Huge Machines. The development of automatic speech recognition (ASR) technology will be presented as a case in point with specific emphasis on efforts to develop ASR technology for the 11 official languages of South Africa during the last 20 years.

Speaker: Dr Febe de Wet (Stellenbosch University)

14:20 Learning image interest and saliency from pairwise image comparisons

Video cameras are increasingly deployed in exploration, monitoring and surveillance applications. These cameras produce vast amounts of information, which needs to be condensed into manageable quantities for both storage and human-operator evaluation. While data compression can address the former, this does not aid operators, who are often faced with the daunting task of analysing lengthy video sequences. As a result, a system that automatically flags interesting images or information and presents this to an operator in a concise manner is highly desirable. This talk will show how the 'interest' value of an image can be inferred by means of a Bayesian ranking framework operating on pairwise image comparisons obtained from users. These interest values can then be used to automatically flag images likely to be of use to a human operator. Algorithm modifications showing how these interest posteriors can be improved by applying Gaussian process smoothing to convolutional neural network feature representations (general datasets) will be discussed and storyboarding applications introduced. The talk will also show how image saliency maps highlighting content of interest to a user can be produced as a byproduct of the interest inference process.

Speaker: Dr Michael Burke (Council for scientific and industrial research/ University of Witwatersrand)

14:40 Machine Learning Based Bitcoin Price Prediction

Blockchain based cryptocurrencies are proving themselves to be both popular and unpredictable. In this project we attempt to design a software system to predict the price movement of Bitcoin using Machine Learning techniques. We focus on feature selection, data processing and the training of various classifier neural networks.

Speaker: Mr Matthew Finlayson (Student)

13:30 → 15:00 Earth Science: III

Convener: Dr Mary-Jane Bopape (South African Weather Service)

13:30 Numerical modeling of transients in the solar modulation of astroparticles

A three-dimensional (3D), time-dependent numerical model for the solar modulation of astroparticles (cosmic rays) is constructed based on solving the relevant transport equation applicable to the heliosphere which is the electrodynamic influence sphere of the Sun. The model is using the Stochastic Differential Equation (SDE) approach which provides numerical advantages. The model is designed to simulate transient events which originate from the Sun and propagate outwards with the solar wind to produce sharp and relative short term decrease, called Forbush Deceases (FDs), in the intensity of astroparticles. These charged particles with energies above 1 MeV arrive at the Earth from Galactic space (cosmos). In order to study the main features of these FDs, a complicated diffusion barrier must be constructed taking into account all relevant physics, such as convection, diffusion, adiabatic energy loses and particle drifts. The model and results will be discussed.

Speaker: Prof. Marius Potgieter (North-West University)

14:00 Detailed projections of future climate change over Africa: the Variable Resolution Earth System Model and supercomputing on the CHPC Lengau cluster

Global Climate Models (GCMs) have become the main tools to project future climate change. GCMs that are capable of also simulating the global carbon cycle are referred to as Earth System Models (ESMs). Currently, about 30 GCMs exist globally, but only one of these (developed by the CSIRO in Australia) had its origin in the southern hemisphere. Moreover, Africa is the only continent for which GCM projections have not improved moving from Assessment Report Four to Assessment Report Five (AR5) of the Intergovernmental Panel on Climate Change (IPCC), indicating a lack of focus on African climate issues in GCM development. Against this background, a new coupled climate model, the Variable-resolution Earth System Model (VrESM) is currently under development through collaborative research between the Council for Industrial Research (CSIR) in South Africa and the Commonwealth Scientific and Industrial Research Organisation (CSIRO) In Australia. The model employs ocean, atmosphere and land-surface models all cast on a cube-based grid and can be applied at quasi-uniform horizontal resolution to function as a global climate model, or in stretched-grid mode to function as a high-resolution regional climate model. The atmospheric model component is the Variable-cubic Atmospheric Model (VCAM) of CSIRO, which has evolved from the widely used Conformal-cubic Atmospheric Model (CCAM). The CSIRO Atmosphere Biosphere Land Exchange model (CABLE) provides VrESM with a dynamic land-surface, whilst the ocean is simulated by the newly developed Variable-cubic Ocean Model (VCOM) of the CSIR. VRESM development is taking place with the immediate objective of generating African-based projections of future global climate change as a contribution to CMIP6, whilst the stretched-grid version of the model and its component models (e.g. CCAM) is to be used to generate simulations of the coupled southern African climate system as part of the second phase of the Coordinated Regional Downscaling Experiment (CORDEX). Here we present an overview of the numerical architecture and the main component models of VrESM. We proceed to showcase the most detailed projections of future climate change ever obtained for the southern African region. These simulations were performed at a resolution of about 8 km in the horizontal for the period 1961-2100 under both low and high mitigation scenarios. For each scenario, six GCMs that contributed to AR5 of the IPCC were first downscaled to 50 km resolution globally using CCAM - as a contribution to CORDEX, and subsequently to 8 km resolution over Cape Town. The projections are analysed with a focus on the current drought in the Cape Town region - insights are provided on whether this drought can be attributed to climate change and on whether more such droughts can be expected under climate change. All VrESM development, and the application of VrESM and its component models, are taking place on the Lengau cluster of the CHPC. The talk thus also provides insights into the computational requirements of performing the high-resolution simulations described above, including the characteristics of the code in terms of scaling on Lengau.

Speaker: Prof. Francois Engelbrecht (CSIR)

14:20 Evaluation of Soil Moisture in the Coupled CCAM-CABLE Model over South Africa Using In-Situ and ESA CCI Satellite Observations

This study evaluates the performance of the coupled CCAM-CABLE model estimates of soil moisture against in-situ measurements from the Skukuza flux tower and ESA CCI satellite observations both at a point and regional scale. In situ measurements are used to evaluate the accuracy of the soil moisture products at a point (location of the flux tower) through the use of conventional statistical measures such as correlation coefficient, normalised mean bias, root mean square error (RMSE) and coefficient of efficiency. The coupled CCAM-CABLE simulations performed best compared with observations for the surface soil moisture and root zone soil moisture for both the dry (March to October) and wet (November to April) seasons. There is evidence of overestimation of observed soil moisture by the various products. Differences in model performance at the surface and the root zone were observed. Following the evaluation at the site level, we further evaluated the performance of CCAM-CABLE, ESA CCI and GLEAM model projections over a grid covering parts of South Africa using correlation coefficient and root mean square difference (RMSD). There is disagreement (i.e. strong negative correlation) between CCAM-CABLE and ESA-CCI mainly during the wet season especially for the ESA-passive product. However, there is moderate to strong correlation between ESA CCI and CCAM-CABLE during the dry period. Similarly, there is evidence of agreement between the GLEAM models and CCAM-CABLE especially during the dry period for most of the grid. The models generally agree on the estimates of root zone soil moisture with correlation coefficients ranging between 0.7 and 1, and RMSD between 0 and 20% for savanna areas, and correlation coefficient between 0.2 to 0.5, and RMSD of 20 to 40% for the grassland. We conclude that the soil moisture estimated using the coupled CCAM-CABLE is comparable with satellite and in situ observations at different soil depths and scales.

Speaker: Mr Floyd Khosa (CSIR)

14:40 A high performance computing approach to cosmic-ray modulation.

The transport of cosmic-rays highly energetic charged particles originating from astrophysical sources, in the heliosphere, the region dominated by plasma of solar origin, can be modelled using the Parker transport equation (TPE). We solve this equation using a 3D time-dependent stochastic numerical solver, taking into account the various processes that modulate the intensities of these cosmic rays on their way to the inner heliosphere and thus to Earth. Where they can pose significant radiation hazard to, e.g. astronauts. This numerical technique lends itself to large-scale parallel computing. Preliminary results of cosmic-rays intensities computed using the CHPC cluster will be presented.

Speaker: Mr Katlego Moloto (NWU)

13:30 → 15:00 HPC: II: Technology

Convener: Dr Milfeld Kent (TACC)

13:30 Enabling The Era of Data-Centric HPC

The latest revolution in High Performance Computing and Artificial Intelligence is the move to a Data-Centric architecture, a collaborative effort among industry, academia, and manufacturers to reach Exascale performance by taking a holistic system-level approach to fundamental performance improvements. Data-Centric architecture exploits system efficiency and optimizes performance by creating synergies between the hardware and the software and between different hardware components. The boundaries between compute, interconnect and storage have change, in order to enable data analysis everywhere. The s

Speaker: Mr Yossi Avni (Mellanox Technologies)

14:00 Striving to the End-User HPC Sweet Spot: Altair Perspective

HPC technology is advancing more rapidly than at any time in the history of computing. Exa-scale compute frameworks will start to emerge in the coming years and will offer enormous raw compute power. Coupled with this, various commercial vendors are developing user-centric HPC business models in the cloud to drive access to these capabilities. The potential of this technology to help solve real-world problems and advance our understanding of the physical world that we live in is clear. As HPC technologists we should ask ourselves the question: How can we make HPC technology ubiquitous in the general computing frameworks of end-users of the technology?

Speaker: Mr Ernst Burger (Altair Engineering S.A. (Pty) Ltd)

14:30 Trends in systems and how to get efficient performance

With all the advances in massively parallel and multi-core computing with CPUs and accelerators it is often overlooked whether the computational work is being done in an efficient manner. This efficiency is largely being determined at the application level and therefore puts the responsibility of sustaining a certain performance trajectory into the hands of the user. It is observed that the adoption rate of new hardware capabilities is decreasing and lead to a feeling of diminishing returns. This presentation shows the well-known laws of parallel performance from the perspective of a system builder. It also covers through the use of real case studies, examples of how to program for energy efficient parallel application performance.

Speaker: Mr Martin Hilgeman (Dell EMC)

13:30 → 15:00 SA NREN: II

Convener: Dr Zukisani Makalima (DIRISA)

13:30 Testing and Training Cybersecurity using a virtual platform

Cybercrime and cyber vulnerabilities are increasing both in number and complexity and pressure is on business and industry to employ cybersecurity professionals that are able to defend systems and infrastructure against cyber-attacks. As a country’s critical infrastructure can be exposed through integration of IoT and Smart Cities, the need has been identified to develop a platform to conduct various tests focusing on securing the implementations before deployment. The platform provides a virtualised space for the simulation or emulation of enterprise networks to conduct various forms of testing as well as the training of cybersecurity specialists. This hyper-realistic simulation virtual training environment normally resembles the enterprise network or operational environments where the cyber workforce can do using red and blue team exercises that can develop their skills to defend their company networks. Doing training on these platform offer the opportunity for a cyber team to work together, as one team, across multiple security domains to defend their networks, reinforcing the skills and opportunity for shared security responsibility for the company, agency or service. In addition it improve their ability to identify advanced attacks. The Network Emulation and Simulation Laboratory (NESL) at the CSIR, is developed to support higher education and industry to perform network bandwidth and performance modelling, cybersecurity training, device research and advanced analytics to study cyber risks and to deliver effective and practical security solutions, all within the safety of an isolated environment. This platform can also be used to do quantitative, qualitative and realistic assessment of potentially ground-breaking cyber technologies for research and development. It also includes hardware in the loop and malware capability for the emulation of real attacks.

Speakers: Mr Aubrey Labuschagne (CSIR), Prof. Joey Jansen van Vuuren (CSIR)

14:00 Your DNS, My DNS - National risk and DNS infrastructure

The Domain Name System (DNS) is critical to the operation of the majority of services on the Internet at large. Many organisations however pay relatively little attention to this foundational protocol. Whether this be in terms of the operational aspects of the DNS infrastructure, or even the choice of registrars and domain name servers. This talk explores the analysis of several snapshots of one of the internet TLD zones, and the subsequent identification of 'hotspots' within the system which serve as potential areas of weakness. Attacks against these identified infrastructural points, whether by malicious actions such as seen in 2016 with the Marai botnet or by national interests in which those systems reside, could have far reaching impacts on network infrastructure. A number of recommendations are put forward as to how to improve the state of DNS security. In particular a checklist for organisations to consider when considering operational risk in operating DNS.

Speaker: Prof. Barry Irwin (Rhodes University)

14:30 Innovation in the Network and Security Space

On Tuesday 5 December, the South African NREN has invited Mr A Wilson of LucidView to discuss the innovative work that we has been doing in the network space. Mr Wilson will be providing an oral presentation on our recent innovations in this arena.

Speaker: Mr Andrew Wilson (LucidView)

13:30 → 15:00 SADC Meeting: IV

SADC Cyberinfrastructure.

15:00 → 15:30 Break

15:00 → 15:30 Poster

Students will be available to answer questions about their posters.

Convener: Dr Andrew Gill (CHPC)

15:30 → 17:00 CHPC PIs & Users BoF

Birds of a feather session for all CHPC principal investigators and users.

Convener: Dr Werner Janse Van Rensburg (CHPC)

15:30 → 17:00 HPC Ecosystems BoF

Birds of a feather session for the HPC Ecosystems sites.

Convener: Mr Bryan Johnston (CHPC Slave)

15:30 → 16:30 SADC Meeting: V

SADC Cyberinfrastructure.

Wednesday, 6 December

08:00 → 09:00 Registration

09:00 → 10:30 Plenary

Convener: Dr Vahed Anwar (CSIR)

09:00 Progress on The SADC Cyberinfrastructure Framework: Supporting Open Data and Open Science in the region

There is a convergence of data and compute intensive paradigms and a proliferation of research and projects in these areas and applications. There also is a global trend in developing distributed federated high performance data cyberinfrastructures and e-infrastructures to support to research, encourage collaborations and to facilitate data sharing for open science. These cyberinfrastructures can also be used to share expensive instruments and resources including high performance computing (HPC),radio telescopes, genomics databases, electron scanning microscopes etc. through connected research networks. These efforts are typically aimed at accelerating research and discovery across domains and countries often reinforcing National, Regional or Continental objectives or geopolitical objectives such as integration. Regionally, the Southern African Development Community (SADC) countries have engaged and developed a regional collaborative Cyberinfrastructure Framework. The vision is of: *An education, research and innovation environment that provides for human capital development and shared access to unique or distributed facilities to impact socioeconomic development in the SADC region and promote knowledge based economy.*

Speaker: Dr Tshiamo Motshegwa (University Of Botswana)

09:45 The three big challenges scientists are confronted with in today’s data driven research – data, intelligence and interaction.

Please join Rüdiger Dorn, Director of Cloud Strategy in the office of the CTO of Microsoft, where he will highlight trends and examples around how compute infrastructures can be scaled, latency can be reduced and the computational throughput can be multiplied. How the arrival of AI and ML allows researchers to tap into toolkits that can augment their work in the area of cognitive services, analytics and machine learning. Lastly with ever increasing data volumes and higher complexity of models, how there is a need to re-think how to visualize, interpret and collaborate on scientific insights, and how new technical opportunities like holograms and augmented reality can help?

Speaker: Mr Kiernan Mike (Microsoft)

10:30 → 11:00 Break

10:30 → 11:00 Poster

Students will be available to answer questions about their posters.

Convener: Dr Andrew Gill (CHPC)

11:00 → 12:30 BioInformatics: I

Convener: Prof. Fourie Joubert (University of Pretoria)

11:00 H3ABioNet –enabling bioinformatics and big data research in Africa

Bioinformatics, or computational biology, is the application of computing to the analysis and interpretation of biological data. With the emergence and reducing cost of new laboratory technologies for high throughput data generation, African scientists are increasingly generating bigger and more complex datasets. This has led to significant challenges in all aspects of data management, including transfer, storage, analysis and interpretation. One of the initiatives which aims to address these challenges is H3ABioNet, a Pan-African bioinformatics network established to build the capacity for large scale genomics research on the continent. A major project within H3ABioNet has been the design of a new genotyping array for African populations, which identifies genetic differences between individuals that may be causative of specific diseases. The design of the array required significant computing hardware at several different stages due to the size and complexity of the data and algorithms. The project incorporated over 4000 human genome sequences, with collectively over 100 million genetics changes (Single nucleotide variants), that needed to be analysed by multiple different software tools, with millions of statistical tests. Through necessity, the H3Africa array design project utilized the CHPC resources for data transfer, storage and processing. This talk will outline the need for high performance computing in bioinformatics, using the array design project as an example, and will also discuss the challenges in big data in the biomedical sciences in Africa and some of the solutions implemented through H3ABioNet.

Speaker: Prof. Nicola Mulder (University of Cape Town)

11:30 Bioinformatics gone Wild!

The National Zoological Gardens of South Africa (NZG) has a strong research component focused on conservation of Wildlife. This is achieved via the Research and Scientific Services Department. In keeping with the times, and considering ongoing advances in newer research technologies, the NZG has moved toward greater implementation of aspects such as Next Generation Sequencing and the Bioinformatics analysis thereof in its research. As such our research currently includes (a) small genome and organelle sequencing assembly and annotation (b) microsatellite development from NGS data (c) microbiome analysis, as well as (d) genotyping by sequencing, all in a wildlife context. While the organization has acquired access to some resources for the analysis of these large datasets, there is certainly a requirement for access to additional resources capable of handling even larger datasets as well as multiple datasets simultaneously. Via a partnership with the CHPC the NZG is therefore currently in the process of optimizing and implementing workflows which address their big data analysis needs. The presentation aims to outline the research activities of the various groups within NZG, in the NGS-Bioinformatics context, along with the resource requirements that the CHPC would supply, in order to ensure the successful implementation of this exciting partnership.

Speaker: Dr Morne Du Plessis (National Zoological Gardens of South Africa)

12:00 Investigation of the interaction between selected divalent cations and the novel Schistosoma mansoni Universal stress protein G4LZI3.

Schistosomiasis, despite several eradication and elimination attempts, still represents one of the most debilitating parasitic diseases of poverty amongst all the Neglected Tropical Diseases. Economically challenged countries bear the full brunt of the disease, coupled with those of other diseases of public health importance such as malaria, HIV/AIDS and tuberculosis. Praziquantel (PZQ), a first-line treatment drug has been in use for the last three decades, but has recently showed traces of drug resistance, hence alternative treatment regimen becomes a priority. Universal stress proteins have been postulated to be upregulated in the Schistosoma worm in response to hazardous environmental conditions. More so, three divalent cations have been postulated to play a role in the mechanism of action of PZQ in bringing about therapeutic response against schistosomiasis. Therefore, this study was aimed at investigating the interaction between these divalent cations and a novel Usp G4LZI3 protein that has been hypothesized as a possible vaccine candidate. Bioinformatics was used to predict the secondary structure of the protein, and to generate a 3D model of the protein before the cations were docked against the G4LZI3 protein. These in silico results were then validated using Isothermal Titration Calorimetry, by titrating the divalent cations against purified and concentrated fractions of the protein. Preliminary results confirmed Ca2+, Mg2+ and Zn2+ as putative ligands and additionally identified ATP and AMP as possible interacting molecules. However, biophysical studies only showed interaction between Mg2+ and the novel G4LZI3 protein. These results provide prospects for future studies towards anti-schistosomal drug development.

Speaker: Ms Priscilla Masamba (University of Zululand)

11:00 → 12:30 Computational Mechanics: I

Convener: Dr Muaaz Bhamjee (University of Johannesburg)

11:00 Computational Fluid Dynamics in Renewable Energy and High-Speed Transport

Speaker: Prof. Ken Craig (Professor)

11:30 Performance Analysis of an HPC Implementation for Large-Scale Propagation Prediction

The concept of Dynamic Spectrum Access (DSA), allows under-utilised radio frequency (RF) spectrum to be made available for Secondary Users (SUs) when not utilised by the Primary User (PU). A geolocation database is currently the preferred approach by regulators to deploy DSA in the TV-bands, i.e. TV white space networks. Central to the geolocation database is the computation node, responsible for performing RF propagation predictions over a large area. RF field strengths are computed for all PU transmitters to determine the geographic location of spectrum available to SUs. RF propagation prediction is computationally intensive and time-consuming, effectively becoming the limiting factor in extending the geolocation based approach for DSA to other parts of the spectrum. We present a High-Performance Cluster Computer (HPCC) implementation for the computation node. Four propagation models are implemented and adapted for parallelisation. Our performance analysis indicates computational speed-up for all propagation models on the HPCC and factors affecting linear scalability are identified. The HPCC implementation is used to study the impact that the choice of propagation model has on spectrum availability to SU’s.

Speaker: Mr Johan Havenga (North-West University)

Slides Havenga_JPH_CHPC_Conference.pdf

11:50 Computational Aeroacoustic Modelling Using Hybrid RANS/LES Methods With Modified Acoustic Analogies

This study considers a numerical approach to identifying noise mechanisms in tandem cylinders to understand aircraft landing gear as a primary contributor to airframe noise during approach and landing. Fluctuations in the flow properties induced by turbulence are computed as well as the corresponding propagations. A hybrid IDDES turbulence model is employed, to compute the boundary layer and fluctuations in the flow properties. Larsson et al. modified Curle’s analogy leading to the derivation of a version of Curle’s analogy that makes use of strictly time derivatives which has been proven to be less sensitive to numerical errors. Brentner and Farassat derived a formulation of the Ffowcs-Williams and Hawkings analogy for a permeable surface enclosing the acoustic sources which accounts for the quadrupole acoustic sources in the flow without the costly calculation of a volume integral. This study will consider the impact of neglecting the volume sources through a comparison of the two modified versions of Curle’s and FWH analogies with the results of other CFD practitioners as well as experimental data.

Speaker: Mr Zamashobane Zama (University of Cape Town)

11:00 → 12:30 DIRISA: I

Convener: Ms Nobubele Shozi (CSIR)

11:00 Best Practices… No, Good Enough Practices, for Building an eResearch enabled Institution - The NWU eResearch Story

What is eResearch? Depending on the point of view, eResearch may be defined as heavily reliant on cyber infrastructure and data science skills or as research data management, repositories, and open science policies, or actually anything in between. Institutions embarking on the eResearch journey should not underestimate the value (both in terms of time and money) that can be gained by first defining what the term means in their own context, before diving in to establish an institutional eResearch programme. “eResearch”, “eScience”, and “cyber infrastructure” are terms that have become popular in recent years mostly due to the change in research that is caused by technology advancement: faster internet penetrating more communities; cheaper data storage, memory and processing power; higher resolution microscopes and cameras; high-throughput research equipment, and more. Various national initiatives attempt to address eResearch needs, including the Centre for High Performance Computing (CHPC), the Data Intensive Research Initiative of South Africa (DIRISA), and the South African Research Infrastructure Roadmap (SARIR) with its various interdisciplinary projects to name a few. These national projects complements resources and programmes available at institutional level. For a researcher as well as research support staff to be successful, it is critical to be aware of changing research practices and funder requirements as well as resources and opportunities available at the various levels - institutional, regional, national, and international. eResearch initiatives have started to take off at a number of academic institutions across South Africa in the past few years. At North-West University (NWU) the eResearch Initiative officially commenced in July 2015 and was driven from the IT director’s office with support from the DVC Research, Technology and Innovation. Stakeholders like the Research Support Office, the Libraries, all faculties and research entities were invited to collaborate from the outset and have played a critical role in the success of the initiative thus far. The NWU eResearch Initiative is characterised by (i) broad participation across both support and academic environments, (ii) involvement from all career stages - from undergraduates to National Research Foundation (NRF)-rated researchers, (iii) integration with and leveraging of regional, national, and international communities and initiatives, and (iv) a collaborative culture. A visual aid (included in the abstract and published under an open license) was developed to help communicate the focus and aims of the NWU eResearch Initiative to various stakeholders. The focus areas are: - research data management; - computational and digital research; - open science; - reproducible research; - and funding (alternative sources). Five key activities are associated with each of these focus areas. They include creating awareness, running training events, community development, integration with related initiatives, and coming up with ways to make the various activities sustainable. ![The NWU eResearch Initiative][1] One of the most successful initiatives on which eResearch was built at NWU, is the Carpentries workshops and train-the-trainer events. These workshops have provided an opportunity to simultaneously address at least four out of the five focus areas as well as four of the five key activities per event. Not only have we been able to develop computational capacity through running Carpentry workshops, but more than 25 researchers and students were also trained as instructors who can propagate the information and training they have received. In this presentation we will share lessons learned about building an eResearch enabled institution. The presentation will include recommendations for institutions that are embarking on eResearch initiatives of their own. These recommendations will take into consideration the diminishing amounts of funding that are available at research and higher education institutions to tackle new challenges. [1]: https://ndownloader.figshare.com/files/9524437/preview/9524437/preview.jpg

Speaker: Ms Anelda Van der Walt (NWU/Talarify)

11:30 Libraries enabling Open Data: UCTs new Institutional Data Repository

A presentation on UCT Libraries implementing an Institutional Data Repository (IDR) service for UCT. Included are: 1. an overview of drivers for Open Access Data Science support (Mandates and Policies, as well as benefits), 2. a description of how a suitable mechanism for compliance was identified (in the context of a federated, national approach); 3. a brief discussion of key features of the platform (Figshare for Institutions) in practice (live instance of the UCT IDR), including outreach (naming competition, data drop-in sessions et al); 4. some closing remarks on various considerations (technical, operational, and conceptual).

Speaker: Mr Niklas Zimmer (University of Cape Town)

11:55 Towards A National Research Data Management Strategy for South African Universities

Amidst a growing number of Open Science mandates for data sharing and reuse, South African universities are scrambling to provide services to the research community to facilitate compliance with the various requirements of numerous international funding agencies. The impending burden posed on institutions to fund this requirement was brought sharply into focus with the introduction of a similar mandate by the national research funding agency, the very lifeblood on which the academic enterprise is reliant. This paper will outline a journey of collaborative effort towards the concept of a shared data service in an exceptional paradigm shift enabled by a proposal for a Tier 2 data node in the national cyberinfrastructure. Insights will be shared of a process of converging interest in a national strategy for research data management, led by DIRISA, the Data Intensive Research Initiative of South Africa, a component of the National Integrated Cyberinfrastructure System (NICIS).

Speakers: Dr Dale Peters (UCT), Dr Mark Hahnel (figshare)

12:20 10 Minutes Q&A

11:00 → 12:30 HPC Education BoF: I

Academics from relevant disciplines (Computer Science, Electronic and Computer Engineering, Computational Physics, Applied Mathematics, Data Science, etc.) and others interested in, or involved with, teaching and curriculum development are invited to discuss how core HPC concepts can be integrated into the undergraduate and postgraduate curricula at South African and African universities.

A broad range of topics related to HPC, including use cases, distributed computing, architectures and interconnects, operating systems, parallel programming and algorithms, file systems, debugging, performance scaling, accelerator technologies, benchmarking and visualization will be considered.

This session will focus on the unique challenges facing academics in South Africa and Africa, either integrating HPC topics into existing courses and curricula or developing and implementing new courses and specializations, and the ways in which these challenges can be addressed. The provision of access to high-end HPC compute infrastructure for student practical assignments and projects, along with teaching resources and course content, will be discussed with the goal of significantly lowering the barrier to entry for academics wishing to teach HPC topics. The alignment of the HPC education and human capital development mandates of the CHPC with other strategist initiatives such as the HPC Ecosystems Project (which has placed, and is continuing to place, donated HPC infrastructure at universities throughout Africa), will be investigated, along with other ways in which the CHPC can support academics in this critical endeavor.

Convener: Dr John Poole (CHPC)

11:00 Overview of HPC course and textbook

Prof. Thomas Sterling (Indiana U.) will provide an overview of his HPC course and new textbook. This course, which provides a modern and accessible introduction to HPC, has been developed and thoroughly field-tested at Indiana University and is now being adopted at universities across the US. Significant resources, including 80 hours of lecture videos, course slides, student exercises and project material, will be made available free of charge. Prof. Sterling has a long relationship with CHPC and has been extremely enthusiastic in his support for HPC education in Africa.

11:00 → 12:30 Material Science

Convener: Dr Eric Maluta (University of Venda)

11:00 Limits of Langmuir

Langmuir was awarded the Nobel prize in 1932 for his contribution to surface chemistry, a field of utmost importance technically seeing that over 70% of all chemicals are produced in a surface-catalysed reaction. He proposed a bold theory on the adsorption of molecules at surfaces, on which basis the rate of heterogeneously catalysed reactions are typically described. The fundamental assumptions underlying the Langmuir theory include inter alia neglecting adsorbate-adsorbate interactions and single type of surface (site). These conditions are not necessarily fulfilled when dealing with catalysis by metals, especially when dealing with reactions at industrial conditions (high pressure) catalysed by nano-sized metal crystallites. Theoretical chemistry can give detailed insight in surface chemistry. In particular, the conditions at which adsorbate-adsorbate interactions become important and their consequences can nowadays be easily explored using DFT. In this talk, the adsorbate-adsorbate interactions of species adsorbed on Fe(100) involved in the methanation, and O/OH co-adsorption on Pt(111) and Pt(100)-surfaces will be discussed.

Speaker: Prof. Eric van Steen (Catalysis Insititute, Department of Chemical Engineering, University of Cape Town)

11:30 Validation of the OSCAR-5 Code System against Experimental Data on Fuel Burnup and Material Activation in Research Reactors

The IAEA (International Atomic Energy Agency) Cooperative Research Project (CRP) 2026, entitled "Benchmarks of Computational Tools against Experimental Data on Fuel Burnup and Material Activation for Utilization, Operation and Safety Analysis of Research Reactors", provides a valuable opportunity to validate calculations tools against quality experimental data. As part of this multi-national effort, Necsa (South African Nuclear Energy Corporation) contributed operational and experimental data of the SAFARI-1 research reactor, and also performed simulations for four other reactors from around the world, using both in-house developed and external codes. The OSCAR-5 systems is the latest iteration of the in-house developed code system which currently supports the operation of the SAFARI-1 reactor, and a number of other research reactors around the world. This version for the first time incorporates support for high fidelity modeling, and therefore has a strong HPC component. This talk will give an overview of the calculational system, in particular how and where computational resources are used, and summarize some initial validation results for the five reactors modeled.

Speaker: Dr Francois van Heerden (Necsa)

11:50 ELSI: A unified software interface for Kohn-Sham electronic structure solvers

Solving the electronic structure from a generalized or standard eigenproblem is often the bottleneck in large-scale molecular and materials simulations based on Kohn-Sham density-functional theory. This problem must be addressed by essentially all current electronic structure codes, based on similar matrix expressions, and by high-performance computation. We present a unified software interface, ELSI (elsi-interchange.org), that simplifies the implementation and optimal use of multiple Kohn-Sham electronic structure solvers. ELSI offers reasonable default parameters for the solvers; dynamic switch between solvers within a single calculation; and automatic conversion between input and internal working matrix formats. Comparative benchmarks performed with multiple solvers on distributed memory supercomputing architectures are shown for system sizes up to tens of thousands of atoms.

Speaker: Victor Yu (Duke University)

12:10 Computational and experimental study on heats of adsorption of MBT, MBO and MBI onto pyrite mineral surfaces

The adsorption of heterocyclic collectors such MBT, MBO and MBI on pyrite mineral surface are paramount in order to establish understanding on the floatation reactivity that may be applicable in a wide range of sulphide minerals. This study explores the heats of adsorption of heterocyclic collectors onto pyrite mineral surfaces. The computed heat of adsorption energies of collectors indicated that the order of adsorption decrease as: MBO > MBI > MBT on both (100) and (111) surfaces. In the case of the pyrite (100) surface, the MBT adsorbed through the exocyclic sulphur atom, while MBO and MBI adsorbed through the exocyclic sulphur and N atoms onto the 5-coordinated Fe atom. The pyrite (111) surface was found to be more reactive and different adsorption mechanisms were predicted, where the MBT exocyclic sulphur atom bridged on the 5-coordinated Fe and surface 2-coordinated sulphur atoms, while the N atom adsorbed on the 5-coordinated Fe atom. In the case of MBO and MBI, dissociation of the exocyclic sulphur atom from the collector occurred with subsequent interaction with the surface 5-cordinated Fe and 2-coordinated S atoms. This resulted in the collector carbon atom adsorbing on the surface 2-coordinated S atom, while the N atom adsorbed on the 5-coordinated Fe atom. The microcalorimetry experimental heats of adsorption of MBT, MBO and MBI indicated that MBI had the strongest adsorption and the order of adsorption decrease as: MBI > MBT > MBO.

Speaker: Dr Peace Mkhonto (University of Limpopo (MMC))

12:30 → 13:30 Lunch

13:30 → 15:00 BioInformatics: II

Convener: Prof. Nicola Mulder (University of Cape Town)

13:30 Computation and the future of biological research

Classic biological research where individual molecules or organisms are studied in isolation, is rapidly being complemented by high-throughput highly-parallel research approaches. The implications are profound for established researchers, but even more so for new post-graduate students entering nearly all fields of biology. This is particularly true for the high impact that massive scale genomic studies are having on work in organisms ranging form viruses and bacteria, up to mammals and humans. Highlights of bioinformatics-related work performed at the CHPC will be shown, including the Forest Molecular Biology programme, the Tree Pathology Co-operative programme, the Maize Fungal Pathogen programme, the HIV Computational Immunology programme, the Malaria Transcriptomics programme and the Cancer Genomics programme. This will be combined with some speculation around the future needs of all biological researcher to be able to operate in a computational and particularly a high-performance computational environment.

Speaker: Prof. Fourie Joubert (University of Pretoria)

14:00 An Integrative Approach to Generating a Reference Transcriptome for Sugarcane

The sugarcane industry is a substantial agricultural activity in South Africa, which contributes significantly to the country’s gross domestic product and positively impacts on the lives of over a million people. Economic loss due to *Eldana saccharina*, a lepidopteran stem-borer, is estimated to be R900 million per annum. A better understanding of the molecular mechanisms involved in resistance against eldana would be an important step towards targeted breeding of resistant sugarcane varieties. In a genomic context, sugarcane is classed as an ‘orphan crop’ as no genome assembly is currently available. However, next generation sequencing (NGS) technologies have enabled large scale sequencing of genomes and transcriptomes providing the data needed to uncover these mechanisms. Commercial sugarcane cultivars (*Saccharum* spp. hybrids) have large and complex polyploid genomes. Moreover, it is not known how gene expression is influenced by the multiple alleles created due to the aneuploid and polyploid nature of the genome. Our aim is to harness the wealth of in-house and publicly available genomic and transcriptomic data in order to construct a reference transcriptome for use in sugarcane RNA-seq studies, whilst overcoming the limitations of *de novo* transcriptome assembly with low coverage data. Not only will this project endeavour to uncover the molecular mechanisms underpinning resistance to eldana, but the approach taken should allow the development of a generic and modular analysis pipeline capable of creating reference transcriptomes, in the absence of a reference genome, in other non-model plant species.

Speaker: Ms Robyn Jacob (South African Sugarcane Research Institute)

14:20 In silico identification and structural bioinformatics of druggable protein targets in Schistosoma Species

Schistosomiasis is a debilitating disease caused by a parasitic flatworm found in freshwater. After malaria, this disease is the second most prevalent disease in Africa and is endemic in the tropical and subtropical regions of the world. Morbidity and mortality attributed to this infection are very high with about 240 million people infected, 800 million persons at risk of the infection and at least 280,000 deaths annually. With the exponential increase in prevalence, Praziquantel (PZQ) remains the only effective drug in the antischistosomal arsenal, which is effective, but ineffective against the juvenile worm. More so, resistance to PZQ has been widely reported of late. Therefore, it is of paramount importance to develop effective alternative antischistosomal compounds using bioinformatics based tools utilizing the broad-spectrum therapeutic capabilities of Antimicrobial Peptides (AMPs). These AMPs are essential components of the innate immune system and are responsible for the complete destruction and immunomodulatory effects in the host defence against pathogenic organisms. Twenty putative antischistosomal AMPs were identified using an in silico mathematical algorithm, Hidden Markov Models (HMMER) but of which six were selected based on their E-values for further exploration. Physicochemical parameters of these six AMPs were computed and their 3D structures were determined using the Iterative Threading ASSEmbly Refinement (I-TASSER) server. Subsequently, the STITCH database was queried to retrieve novel Schistosoma proteins that can potentially serve as targets for the identified AMPs. Glycosyltransferase and Axonemal dynein intermediate chain protein were identified as the novel druggable target proteins. Thereafter, the physicochemical characterization and prediction of the 3D structural elements of the aforementioned proteins were carried out as well. Finally, PatchDock was used to dock the 3D structures of the putative antischistosomal AMPs against the 3D structures of the druggable proteins. Overall, TAK3 showed a good binding affinity with glycosyltransferase and TAK6 displayed the highest binding affinity with axonemal dynein intermediate chain.*emphasized text*

Speaker: Mr Raphael Taiwo Aruleba (University of Zululand)

13:30 → 15:00 Chemistry

Convener: Prof. Catharine Esterhuysen (Department of Chemistry and Poymer Science, University of Stellenbosch)

13:30 DFT for isomers in solution, redox potentials, transition states and favored reaction products

Apart from merely computationally running molecular energy optimizations and determining optimal geometries, a large variety of other techniques and applications also exist. In this presentation uncomplicated methods are shared by which: 1. isomer geometries in solution may be determined by involving TDDFT, 2. redox potentials of series of metal complexes are determined merely from optimization HOMO & LUMO energies, 3. transition states of photo-excited species are established from energy minima on computed energy surfaces, and 4. favored complex reaction products are computed by means of reaction energy and molecular orbital calculations. References: 1. Dithizone and its Oxidation Products – a DFT, Spectroscopic and X-Ray Structural Study, Von Eschwege, K. G.; Conradie, J.; Kuhn, A., J. Phys. Chem. A, 115, 14637-14646, 2011, http://pubs.acs.org/doi/pdf/10.1021/jp208212e 2. Redox Potentials of Ligands and Complexes – a DFT Approach, Von Eschwege, K. G.; Conradie, J., SA Journal of Chemistry, 64, 203–209, 2011, http://reference.sabinet.co.za/webx/access/electronic_journals/chem/chem_v64_a33.pdf 3. Ultrafast Photochemistry of Dithizonatophenylmercury(II), Schwoerer, H.; Von Eschwege, K. G.; Bosman, G.; Krok, P.; Conradie, J., ChemPhysChem, 2653-2658, 2011, http://onlinelibrary.wiley.com/doi/10.1002/cphc.201100337/abstract 4. Synthesis and structure of dithizonato complexes of antimony(III), copper(II) and tin(IV), Trevor T. Chiweshe, Marilé Landman, Jeanet Conradie & Karel G. von Eschwege, Journal of Coordination Chemistry, 2016, 69(5) 788–800, http://dx.doi.org/10.1080/00958972.2016.1145213

Speaker: Prof. Karel von Eschwege (University of the Free State)

13:50 FAMSEC-based insight on the chemical nature of the XH...HY (X,Y = C,N) intramolecular interactions

Speaker: Prof. Ignacy Cukrowski (University of Pretoria)

14:10 The degradation of phosphate derived ligands in hydrometallurgical solvent extraction

Speaker: Mr Daniel Jansen van Vuuren (North-West University)

14:30 A DFT study of the ODH of n-hexane over isolated H3VO4 and H4V2O7

Catalytic (H3VO4 and H4V2O7) oxidative dehydrogenation (ODH) mechanistic studies of the reaction of n-hexane to 1- and 2-hexene have been conducted by means of Density Functional Theory (DFT). The aim of this study was to gain insight about the catalytic mechanistic pathways for the conversion of n-hexane to 1- and 2-hexene. The 3-hexene pathway was omitted because the isomer is not observed in our experiments, nor reported in literature. The chosen models for the study were the isolated H3VO4 and H4V2O7 clusters, with the H3VO4 unit comprising one vanadyl bond, V(V)=O, and the H4V2O7 cluster comprising two vanadyl bonds and the bridging O atom in the V(V)-O-V(V) unit. The calculated rate-determining step is β-hydrogen abstraction from the C6H14 chain by vanadyl O to produce the complex intermediates with reduced V(IV) centres in accumulated V(IV)-OH bonds and activation barriers of ΔE# = +27.4 (H3VO4) and +32.7 (H4V2O7) kcal/mol. Both these values are lower than the value calculated for the H-abstraction by the bridging O in H4V2O7 (ΔE# = +43.9 kcal/mol). The energetically favourable propagation steps that may lead to olefins involve α-H abstraction (1-hexene) and γ-hydrogen abstraction (2-hexene) on the radical intermediate fragment (⋅C6H13) by vanadyl O from a different site or by gas-phase molecular O2. The gas-phase pathway may dominate at lower n-hexane to oxygen molar ratios combined with low V(V)=O surface areas and be subdued at higher molar ratios in combination with high V(V)=O surface areas. However, chemisorption of the radical intermediate (∙C6H13) on the surface O sites may lead to undesired products including oxygenates. This may explain the low yields of 1- and 2-hexene (< 20%) obtained in our laboratory experiments. H-transfer between two V(IV) centres (accumulated V(IV)-OH) reproduces V(V)=O, and also V(III) and H2O. The reoxidation of reduced V(III) to V(V) occurs through Mars–van Krevelen mechanism. All structures on the potential energy surfaces (PESs) were optimized using the GAUSSIAN 09W program, at the B3LYP level, with the 6-311+g(d,p) basis set for C, O and H atoms and effective core potentials (ECPs) for the V atom. The laboratory experimental conditions of 573, 673 and 773K were included in the computations. The most energetically favourable pathways for the reaction were determined from some calculated kinetic and thermodynamic properties (ΔE#, ΔE, ΔG# and ΔG) and the catalytic mechanism that is likely to be followed will be discussed.

Speaker: Dr Nkululeko Damoyi (Mangosuthu University of Technology)

13:30 → 15:00 Computational Mechanics: II

Convener: Prof. Ken Craig (Professor)

13:30 Modelling of Multiphase Flow in Process Equipment: The Trade-off between Accuracy and Computational Efficiency

Process equipment behaviour is characterised by multiscale multiphase interactions namely: liquid-gas, fluid-particle, particle-particle and particle-wall interactions. The understanding of these interactions, ideally, requires a general solution to the full turbulent Navier-Stokes equations combined with a unifying theory for granular flow. However, in the absence of such breakthroughs advances in Computational Fluid Dynamics (CFD) and Computational Granular Dynamics (CGD) are essential. It is evident from the literature that significant advances in the theory of fluid and granular dynamics has had a positive bearing on the advancement of the industrial value of chemical, metal and mineral processing industries [1,2,3]. According to Joshi and Ranade [1] CFD is a technology that can be used to achieve these advances from the fluid dynamics perspective. In the same light, Pöschel and Schwager [3] assert that computational granular dynamics (CGD) is also necessary to achieve these advances. Based on the advances in both fields it is clear that a combination of both is required to make the necessary leaps in technology required to achieve the above. However, the combined modelling of the complex physics is often computationally intensive, leaving the practitioner with the trade-off between the accuracy of the model predictions and the computational efficiency of the model. The aim of this research is to improve the understanding of the physics of such flows and to provide accurate and computationally efficient models to be used in design optimisation of mineral processing processes and equipment. Currently hydrocyclones have been used as the examplar. However, broader applications are being introduced. Multiple vartiations of Navier-Stokes and Lattice Boltzmann Method (LBM) based models, coupled with the Discrete Element Method (DEM), are employed in the research. The models are benchmarked against experiment to determine accuracy and also benchmarked against each other for computational efficiency. The major contribution of this research is the demonstration that the LBM can provide predictions of the multiphase flow and interactions in a hydrocyclone that are at least comparable to, and in some cases superior to, the Navier-Stokes based approach whilst remaining more computationally efficient (or comparable) to the Navier-Stokes based approach. Future work will focus on the development of a stable two-fluid LBM model and coupled a LBM-DEM model. [1] J. B. Joshi and V. V. Ranade, “Computational fluid dynamics for designing process equipment: expectations, current status, and path forward,” Industrial & Engineering Chemistry Research, vol. 42, pp. 1115–1128, 2003. [2] H. E. Van den Akker, “Toward a truly multiscale computational strategy for simulating turbulent two-phase flow processes,” Industrial and Engineering Chemistry Research, vol. 49, pp. 10780–10797, 2010. [3] T. Pöschel and T. Schwager, Computational Granular Dynamics: Models and Algorithms, 1st ed. Heidelberg, New York: Springer-Berlin, 2005.

Speaker: Dr Muaaz Bhamjee (University of Johannesburg)

14:00 Computational modelling of furnace tapholes – a case study in life at the interface between academic and industrial research

Pyrometallurgical smelting furnaces are used in the production of many commodities of strategic and economic importance in South Africa including ferrochromium, platinum-group metals, titanium dioxide, manganese alloys, and others. Most furnaces make use of tapholes, channels through the furnace wall and lining, in order to facilitate the removal of molten process material from inside the vessel. Correct understanding and operation of the taphole is essential for optimal performance of such furnaces [1]. Depending on the smelting process being used the furnace may use a single taphole for all material, or separate tapholes for different phases such as slag and metal. Tapholes are typically located in a taphole assembly, a specialised section of the furnace sidewall designed for this purpose. Fluid flow of molten material through furnace tapholes during tapping is a complex problem, and its qualitative and quantitative behaviour is generally governed by a combination of taphole geometry and properties of the tapped fluid [1,2]. Understanding the interaction between these parameters and the operating conditions inside the furnace is important for determination of tapping flowrates and wear phenomena. In this presentation, a technical overview of recent work at Mintek on computational modelling of fluid flow through furnace tapholes will be presented. In addition, the relationship between fundamental research and industrial development in the pyrometallurgical sector will be explored in the context of taphole modelling. In particular, factors which can limit innovation in pyrometallurgy – such as strong risk sensitivity, harsh macro-economic climates, and long lead times for adoption of new technologies – will be addressed. The leveraging of academic research using facilities such as CHPC is seen as a critical and necessary step in the process of engaging with industrial clients and technology partners, especially in the relatively new and disruptive field of computational modelling. 1. Nelson, L.R. and Hundermark, R.J. (2016). ‘The tap-hole’ – key to furnace performance, J. SAIMM, 116(5), p 465 2. Reynolds, Q.G. and Erwee, M.W. (2017). Multiphase fluid flow modelling of furnace tapholes, Proc. CFD 2017, p 521

Speakers: Mr Markus Erwee (Mintek), Dr Quinn Reynolds (Mintek)

14:30 Adjusting empirical relationships in a turbulence model

This talk focuses on the methodology taken in adjusting turbulence model coefficients based on historical data. The problem background as to why the turbulence model needed to be adjusted and how the HPC was able to generate data for surrogate model optimization. The SST transition model as 1st described by Menter in 2001, makes use of empirical relationships to capture the effect of transitional flow in CFD. These relationships were made public in 2006, for a variety of geometries. The determination of these relationships were based on experimental data and aimed to ensure the correct flow effects, but do not capture the physics of the flow. While the model has been in use for some time and proven accurate in both academic and industry applications, the specific relationships remain proprietary in commercial codes (such as ANSYS). The open source software openFOAM was selected to perform CFD simulations in an effort to identify relationships specific to airfoil geometries. Without performing experimental tests on an airfoil, historic experimental data was used as the test case. In particular the NACA0012 foil was selected for its extensive collection of published experimental data.

Speaker: Mr Gareth Erfort (Stellenbosch University)

13:30 → 15:00 DIRISA: II

Convener: Ms Nobubele Shozi (CSIR)

13:30 A next generation institutional repository for ALL South African research

The institutional repository (IR) has played an integral role in the scholarly infrastructure over the last 25 years. As the dissemination of digital information at the institutional level is accelerating and research outputs are diversifying, we find ourselves at a crossroads of the future direction of the IR. To compound the complexity, not only do researchers want to share research in a variety of formats, they want to share it in ways the traditional scholarly publishing model does not cater to. With the rise of new cultures in research, there is an increasing desire to publish results early and often, iterate on research, and conduct peer review in a more transparent way after the point of publication. With this new style of scholarly communication on the horizon, some have signaled the death knell of the IR, however with some technical improvements, we at Figshare see the IR as the natural home for ALL scholarly outputs, including research data. In this talk we will discuss national pilots of figshare infrastructure to allow all South African academics to get credit for their research.

Speaker: Dr Mark Hahnel (figshare)

13:55 What infrastructure is necessary for successful research data management (RDM) at universities?

The presentation will cover the following elements for RDM success: • RDM life cycle • Research data elements in the research life cycle with international examples • What can be seen as RDM infrastructure • IT infrastructure • Library infrastructure • Research office infrastructure • Recommendations and conclusion

Speaker: Dr Heila Pienaar (University of Pretoria)

14:20 Incentives driving data sharing: an African perspective

The Open Science movement – focused on making research data, software code and experimental methods publicly available and transparent - is steadily gaining momentum. According to Gewin (2016), “[A]a spirit of openness is gaining traction in the science community, and is the only way, say advocates, to address a 'crisis' in science whereby too few findings are successfully reproduced. Furthermore, they say, it is the best way for researchers to gather the range of observations that are necessary to speed up discoveries or to identify large-scale trends.” Although many researchers are already sharing their raw data and data sets, there are researchers who still question why they should share their data, and what benefits are in it for them. In other words, how will they be incentivized, when others use the outputs of their hard work. These issues also apply on national and institutional levels, across the world. Very few countries and institutions have policies in place regarding the management (incl. curation) and sharing of data as an outcome of funded research projects. Through this talk, some initial findings on incentives for sharing data from the African Open Science Platform project will be presented.

Speaker: Mrs Ina Smith (Academy of Science of South Africa (ASSAf))

14:45 10 Minutes Q&A

13:30 → 15:00 HPC Education BoF: II

Academics from relevant disciplines (Computer Science, Electronic and Computer Engineering, Computational Physics, Applied Mathematics, Data Science, etc.) and others interested in, or involved with, teaching and curriculum development are invited to discuss how core HPC concepts can be integrated into the undergraduate and postgraduate curricula at South African and African universities.

A broad range of topics related to HPC, including use cases, distributed computing, architectures and interconnects, operating systems, parallel programming and algorithms, file systems, debugging, performance scaling, accelerator technologies, benchmarking and visualization will be considered.

This session will focus on the unique challenges facing academics in South Africa and Africa, either integrating HPC topics into existing courses and curricula or developing and implementing new courses and specializations, and the ways in which these challenges can be addressed. The provision of access to high-end HPC compute infrastructure for student practical assignments and projects, along with teaching resources and course content, will be discussed with the goal of significantly lowering the barrier to entry for academics wishing to teach HPC topics. The alignment of the HPC education and human capital development mandates of the CHPC with other strategist initiatives such as the HPC Ecosystems Project (which has placed, and is continuing to place, donated HPC infrastructure at universities throughout Africa), will be investigated, along with other ways in which the CHPC can support academics in this critical endeavor.

Convener: Dr John Poole (CHPC)

13:30 HPC Topics in African University Courses

This session will look at current efforts and success in include HPC topics in courses at African universities, including teaching strategies and methodologies. The challenges regarding implementation of HPC education in an African context will be discussed as well solutions and the next steps forward. A key goal is building and strengthening a community around HPC education in Africa.

15:00 → 15:30 Break

15:00 → 15:30 Poster

Students will be available to answer questions about their posters.

Convener: Dr Andrew Gill (CHPC)

15:30 → 16:30 DIRISA: Panel Discussion

Convener: Dr Tshiamo Motshegwa (University Of Botswana)

15:30 Panel Discussion: Making Data Sharing Work

Questions: 1. What are the obstacles to data sharing and how can these be addressed? 2. How can we raise awareness among researchers on data sharing? 3. How can data sharing be incentivised? 4. What resources are needed to make data sharing work?

Speaker: Dr Tshiamo Motshegwa (University Of Botswana)

15:30 → 16:30 Women in HPC BoF

Birds of a feather session dedicated to women in HPC.

Convener: Prof. Regina Maphanga (CSIR)

15:30 Women in High Performance Computing in South Africa

The first session on Women in High Performance Computing in South Africa was proposed and held during 2016 annual conference. The major aim of the initiative was to establish a network of Women in HPC in SA, by bringing them together during the meeting. The session was supported and attended by both men and women, and most importantly was supported by CHPC management team. An interim steering committee comprising of women from various institutions and disciplines was established and will take the initiative going forward. Consequently, we propose to have a session during 2017 annual conference. It is envisaged that at the end of the session and beyond, the following goals would be achieved: • Create a long-term platform for interaction and mentoring among the women. • Encourage more female learners and students to consider careers in HPC. • Contribute in increasing the number of women and girls participation in HPC through training and networking. • Share information and resources that foster growth for women in HPC.

Speaker: Regina Maphanga (CSIR)

18:00 → 18:45 Plenary: Plenary 7

Convener: Dr Happy Sithole (CHPC)

18:00 The Dawning of the Neo-Digital Age in the Era of Nano-Scale Technology

This is the Dawning of the Neo-Digital Age; a unique period brought on by the end of Moore’s Law as semiconductor fabrication enters its final phase with nano-scale feature size. The Neo-Digital Age also corresponds to a time when energy constraints at the chip level and the system level bound power consumption and indirectly the clock rate of device logic. It will prove to be an epoch of innovation, in part out of desperation to achieve yet further significance performance gains even with the emerging barriers, but also through an explosion of creativity in computer architecture not seen since the 1980s. Ideas are already percolating and being considered by academia, industry, and even industry who has held fast to approaches based on incremental changes to otherwise conventional methodologies. Such currently pursued strategies include neuromorphic, quantum computing, cellular automata and others. Some of these have their roots in research of prior decades. This plenary presentation will discuss in some detail one possible neo-digital class of computing, the Continuum Computer Architecture, and at greater depth one specific example being explored, the Simultac Fonton. Analysis suggests that the Simultac non-von Neumann architecture using today’s semiconductor technology could deliver a peak exaflops performance at 10% the expected cost of a conventional derivative machine at exascale, 10% power consumption, and 1% size. The Simultac use of dynamic adaptive introspection and the HPX+ runtime system software support based on the ParalleX execution model (previously discussed) will significantly enhance efficiency, scalability, and user productivity. Questions will be encouraged by participants throughout the presentation as well as the Q&A session at the end.

Speaker: Thomas Sterling (Indiana University)

18:45 → 19:45 Plenary: Prize-Giving

Convener: Dr Happy Sithole (CHPC)

18:45 Poster Awards

Speaker: Dr Happy Sithole (CHPC)

18:55 Student Cyber Security Challenge Awards

Speaker: Dr Renier van Heerden (SANReN, CSIR)

19:10 Student Cluster Competition Awards

Speaker: Dr Happy Sithole (CHPC)

19:45 → 21:00 Dinner: Cocktails

Thursday, 7 December

08:00 → 09:00 Registration

09:00 → 12:30 CHPC Industry Engagement Meeting

The CHPC's non-academic users are invited to the annual industry engagement meeting,to be held on the morning of 7 December, with informal discussions to be continued during lunch. The purpose of the meeting is to touch base with existing non-academic users, discuss common points of interest and to inform prospective new commercial users. The meeting will consist of a presentation by CHPC staff, followed by brief presentations by users and a group discussion. The intention is for the CHPC to learn from the users' experience, and to inform the users of future plans.

09:00 CHPC Industry Engagement Meeting

The CHPC's non-academic users are invited to the annual industry engagement meeting,to be held on the morning of 7 December, with informal discussions to be continued during lunch. The purpose of the meeting is to touch base with existing non-academic users, discuss common points of interest and to inform prospective new commercial users. The meeting will consist of a presentation by CHPC staff, followed by brief presentations by users and a group discussion. The intention is for the CHPC to learn from the users' experience, and to inform the users of future plans.

Speaker: Dr Werner Janse Van Rensburg (CHPC)

10:30 Break

09:00 → 17:00 Thursday Workshop: Deep Learning training course: OpenPOWER and AI

his workshop will explore acceleration opportunities in existing and upcoming workflows. We will also explore the creation of a community & ecosystem around POWER9 acceleration technology for academics and industry practitioners.

09:00 OpenPOWER Workshop

This workshop will explore acceleration opportunities in existing and upcoming workflows. We will also explore the creation of a community & ecosystem around POWER9 acceleration technology for academics and industry practitioners.

Speaker: Mr Ganesan Narayanasamy (TECHNOLOGY leader)

10:30 Break

12:30 Lunch

15:00 Break

09:00 → 17:00 Thursday Workshop: HPC configuration management using Puppet 5

In this workshop we will install and configure Puppet 5, a configuration management system, to manage our compute cluster.

We will install and configure the Puppet server, PuppetDB database, a Puppet agent and multiple modules from the Puppet Forge. Best practises will be implemented, including using Hiera and saving configurations to Git.

The workshop will be hands-on. Please ensure you have a working installation of Hyper-V, VMware (Workstation or Player) or VirtualBox. Copies of the VMDK/VHDX files will be provided.

09:00 HPC configuration management using Puppet 5

In this workshop we will install and configure Puppet 5, a configuration management system, to manage our compute cluster. We will install and configure the Puppet server, PuppetDB database, a Puppet agent and multiple modules from the Puppet Forge. Best practises will be implemented, including using Hiera and saving configurations to Git. The workshop will be hands-on. Please ensure you have a working installation of Hyper-V, VMware (Workstation or Player) or VirtualBox. Copies of the VMDK/VHDX files will be provided.

Speaker: Mr Charl Möller (University of Stellenbosch)

notes https://cwmoller.github.io/workshop-puppet5/

10:30 Break

12:30 Lunch

15:00 Break

09:00 → 12:30 Thursday Workshop: Introduction to Parallel Computing with OpenMP

This course will present an introduction to foundational concepts of parallel programming, and will specifically train attendees in the basic use of OpenMP for C/C++ and Fortran programmers. The fundamentals of creating parallel execution blocks, forking and synchronizing threads, and employing worksharing loops will be featured in a lecture and explored in a hands-on lab. Additionally, an introduction to new features in OpenMP (SIMD, Affinity) and a high-level introduction to tasking will be presented.

09:00 Introduction to Parallel Computing with OpenMP

Speaker: Dr Milfeld Kent (TACC)

10:30 Break

09:00 → 12:30 Thursday Workshop: Microsoft AI "Amplifying human ingenuity"

This workshop is for everyone who wants to learn, develop, or super-charge their research with AI and machine learning. You’ll discover Microsoft AI and Machine learning platform offerings . Come along if you want to experience hands-on deep-learning for image recognition, natural language processing, speech recognition, deploying machine learning apps using R and Python, mining academic publications, analyzing IoT data, or want to ask questions about how AI can help your research.

10:30 Break

12:30 → 13:30 Lunch

12:30 → 13:30 Lunch break

13:30 → 17:00 Thursday Workshop: Basic Molecular Dynamics Using DL_POLY

This one day workshop is intended for undergraduate project students, postgraduate students, postdoctoral researchers and researchers who are familiar with the field of Molecular Dynamics (MD) and want to employ state-of-the art methodology based on the density functional theory to understand bulk materials properties, surface science and heterogeneous catalysis phenomena. Molecular dynamics is a computer simulation method for studying the physical movements of atoms and molecules. The MD method can assist one in obtaining the static quantities and dynamic quantities.

13:30 Setting Up Basic Molecular Dynamics (MD) Calculations at CHPC Using DL_POLY Code”

Proposal for a Tutorial at 2017 CHPC National Meeting Title: “Setting Up Basic Molecular Dynamics (MD) Calculations at CHPC Using DL_POLY Code” Lecturer(s): 1. Sylvia Ledwaba, University of Limpopo, raesibe.ledwaba@ul.ac.za 2. Cliffton Masedi, University of Limpopo/CSIR, cmasedi@csir.co.za Description: Molecular dynamics (MD) is a computer simulation method for studying the physical movements of atoms and molecules. The MD method can assist one in obtaining the static quantities and dynamic quantities. This method gives a route to dynamical properties of the system: transport coefficients, time-dependent responses to perturbations, rheological properties and spectra. The atoms and molecules are allowed to interact for a fixed period of time, giving a view of the dynamic evolution of the system. The DL_POLY Code parallel molecular dynamics simulation package will be utilised for exploration of such properties of molecular systems. Duration: 1 day (6 hours) Size: (max number of available seats) Target Audience: This one day workshop is intended for undergraduate project students, postgraduate students, postdoctoral researchers and researchers who are familiar with the field and want to employ state-of-the art methodology based on the density functional theory to understand bulk materials properties, surface science and heterogeneous catalysis phenomena. Prerequisites: Familiarity with UNIX or Linux environment. Type of tutorial: Mix of tutorials and hands-on (mostly practical) Special requirements: Attendees need access to a laptop or workstation, active CHPC user account cluster with access to DL_POLY software. Open source spreadsheet such as Microsoft Excel with graphing capabilities is required for post-processing of output data. Free visualization softwares (e.g. VMD, VESTA, etc.), necessary for structure visualization. Outline of full syllabus: Introduction to Molecular Dynamics Properties of Molecular Dynamics Simulation code (DL_POLY) Fitting and validation of interatomic potentials Setting up molecular dynamic simulations Force Field Models/ Interatomic Potentials General Description Potential Models (Buckingham + Three-Body) Compatibility of Interatomic Potentials Validation Hands-on exercises on: Construction of Structure File Validation of Interatomic Potentials Generation of DL_POLY input files Submission of Calculations at CHPC_Lengau

Speakers: Mr Cliffton Masedi (CSIR/CHPC), Dr Raesibe Sylvia Ledwaba (University of Limpopo)

15:00 Break

13:30 → 16:30 Thursday Workshop: CHPC Induction Course

The induction course covers the use of the CHPC systems from the user's point of view.

13:30 CHPC Induction Mini-course

Speaker: Mr Kevin Colville (CHPC)

14:50 Break

13:30 → 17:00 Thursday Workshop: Using the open source CFD GUI in FreeCAD

13:30 Using the Open Source CFD GUI in FreeCAD

The computational fluid dynamics (CFD) workbench for FreeCAD was developed to provide South African industry and academia with a cost-effective solution for modelling fluid flow. This workbench aims to help users set up and run CFD analyses. It guides the user in selecting the relevant physics, specifying the material properties, generating a mesh, assigning boundary conditions and setting the solver settings before running the simulation. Where possible, best practices are included to improve the stability of the solvers. During the workshop, we will give an overview of the technology and demonstrate how to: * Create the geometry; * Selection of simulation parameters; * Specify boundary conditions; * Unstructured mesh generation; * Initiate the run and monitor residuals; and * Visualise the results The CFD workbench is available as an add-on to FreeCAD while more information on the workbench as well as, additional, training material are available on the OpenSim community website: https://opensimsa.github.io/

Speaker: Dr Johan Heyns (CSIR)

15:00 Break