High Performance Computing in South Africa: Computing in Support of African Development
Rob Adam, Director General's Office - Department of Science and Technology, Pretoria
Cheryl de la Rey, Deputy Vice-Chancellor's Office, Research & Innovation - University of Cape Town
Kevin J. Naidoo, Computational Chemistry, Department of Chemistry - University of Cape Town
Daya Reddy, Centre for Research in Computational & Applied Mechanics - University of Cape Town
CTWatch Quarterly
February 2006

1. Introduction

In his 2002 State of the Nation address, President Thabo Mbeki of South Africa singled out Information and Communication Technology (ICT) as "a critical and pervasive element in economic development," and recommended the establishment of an "ICT University." The National Research and Development Strategy of South Africa had earlier also clearly identified ICT as one of the key technology platforms of the modern age, and one which has a central place in initiatives aimed at promoting development in South Africa.

The vision presented by President Mbeki has taken concrete form in the establishment of the Meraka Institute, the purpose of which is to facilitate national economic and social development through human resource development and needs-based research and innovation, leading in turn to products, expertise and services related to Information and Communication Technologies (ICT).

The Centre for High Performance Computing (CHPC)1 is a component of the Meraka Institute. This article describes the objectives and structure of the CHPC and the progress that has been made to date in the establishment of this facility.

2. Background

South Africa is currently in the throes of expanding its scientific research and innovation base and is at the same time identifying focal directions, many of which have a direct link to social and economic development . While the National R&D Strategy sets the framework, there was a recognition that an ICT strategy was needed to chart a comprehensive national approach to ICT R&D in order to maximise its potential economic contribution. Through a co-ordinated national approach, a country like South Africa could not only develop the critical mass to boost it own national development, but also achieve international competitiveness in identified focal areas.

The overall purpose of the national ICT Strategy is to create an enabling environment for the advancement of ICT R&D and Innovation in identified domains. Computational Science and High Performance Computing are two of these. This stems from the firm recognition that access to high performance computing facilities is of central importance to the success of the technology missions identified in the National R&D Strategy. Key examples in this regard are Biotechnology, particularly with reference to research into the major infectious diseases such as HIV/AIDS and tuberculosis, advanced manufacturing technology (e.g., computational simulations of design and manufacturing processes, and computational materials design), technologies to utilise and protect our natural resources and ensure food security (e.g., climate systems analysis and disaster forecasting), and technology for poverty reduction (e.g., behavioural modelling in social research; financial management; HPC in SMEs). Similarly, a number of science missions were identified in the R&D Strategy as standing to benefit from the establishment of an HPC; examples are the Square Kilometre Array (SKA), the National Bioinformatics Network (NBN) and Global Earth Observing System of Systems (GEOSS). High Performance Computing is therefore clearly perceived, in relevant national strategic plans, to be a platform for scientific and technological innovation through which the national R&D strategy can be accelerated. The dual impact of such a platform will be increased global competitiveness and improved local quality of life.

Funding for three years (2006-2008) has been secured for the high performance computing initiative. In addition, parallel investment in a South African National Research Network (SANReN), intended to provide high bandwidth connectivity for South African researchers, has been planned.

3. An African Renaissance in Technology and Development

The developments within South Africa are aligned with initiatives to stimulate research, development and technology across the African continent. A 'Plan for Collective Action' was adopted by African Ministers of Science and Technology in Dakar in November 2005, in a meeting organized jointly by New Partnership for Africa's Development (Nepad)2 and the African Union (AU). This Plan was developed 'bottom-up' by engaging in consultations with scientists and institutions cross all five regions of the continent. It lays out programmatic initiatives and projects that are crucial to enable Africa to mobilize and strengthen its capacities to engage effectively in scientific and technological development. The Plan contains concrete actions that will build the continent's research base and stimulate innovations to fight poverty, increase economic competitiveness and promote human development in general. The Plan complements a series of AU and NEPAD programmes for such areas as agriculture, environment, infrastructure, industrialization, education and energy.

The three conceptual pillars of the 'Plan for Collective Action' are capacity building, knowledge production, and technological innovation. The Plan has twelve sub-programmes based on specific content areas, one of which is Information and Communications Technology. The ICT sub-programme will aim at establishing a continental research network on ICTs. It will bring together leading universities and research centres to design and implement projects that generate software and use with African content. Its specific goals will be to:

Currently, a significant retarding factor is the exorbitant price of bandwidth on the African continent. Fortunately steps are now being taken to address this, with particular attention being given to reasonably priced connections to Europe and from there to other continents. Multinational negotiations regarding the laying of a cable up the east coast of Africa (the so-called Eassy cable) are far advanced.

4. The ICT Roadmap and the Meraka Institute3

Following President Mbeki's directive, the Meraka Institute was launched in May 2005. This is a public/private partnership seeking to promote co-operation between universities, industry and government on ICT learning, research, development and innovation informed by practical, needs-based challenges. The mandate of the Institute, which has been established as a national research centre within the Council for Scientific and Industrial Research, is to:

  1. undertake world-class, needs-based basic research in the ICT field leading to development and innovation to the benefit of South Africa and the region.
  2. develop ICT knowledge workers with sound qualifications.
  3. establish SA as a highly competent international ICT player.
  4. attract leading ICT knowledge workers from various parts of the world.
  5. be the champion, voice, parent and mentor of an emergent South African ICT industry that is regionally relevant and globally competitive.

The Meraka Institute is guided by the fundamental principle that people are the prime basis for success in this knowledge intensive area. To date a number of areas of competency have been earmarked for inclusion in the Meraka Insititute such as ICT for Disability, Human Language Technology, and an Open Source Centre. The CHPC is a major division within the Meraka Institute.

5. Centre for High Performance Computing (CHPC)

Global trends in the development of CHPCs have given strong impetus to plans to establish a CHPC in South Africa. The CHPC will function as a major national innovation platform, which is set to deliver a significant return on investment for the country, by

Figure 1

Figure 1. Core business concept of the CHPC.

The core CHPC facility will be established in Cape Town, with the University of Cape Town designated as its formal host. The considerable existing strengths in scientific computing at various research institutes in South Africa form a large pool of expertise on which the CHPC will be based. Indeed, planning for the establishment of a CHPC began in Western Cape universities with a group of researchers in areas such as computational chemistry, grid computing in theoretical physics, climate modelling, bioinformatics, computational mechanics, radar signal processing and machine vision. Since assuming national proportions, the pool of expertise in scientific computing and HPC has been expanded, for example to include groups in computational physics at universities in Limpopo and Kwazulu-Natal provinces.

To ensure the development of expertise in a truly national fashion and to mitigate the drawbacks of a centrally located national facility, it is planned to establish regional nodes at a number of centres in South Africa. Far from being clones of the central CHPC, these will typically comprise small clusters, with a capacity of not greater than 1/20 of that of the parent facility, which will permit at a local level the initiation of research projects with an HPC dimension, as well as facilitating education and training in HPC. Access to the main CHPC is nevertheless a central plank in the planning, and regional centres will simultaneously have remote access to the central facility.

National and International bandwidth

Although the CHPC will provide an HPC environment that is fully functional as a stand-alone unit, connectivity of the national research institutions to the CHPC, as well as international connectivity to global networks, will have a significant impact on its role as a national resource and ease of access, as well as its reputation as a global player in the field of HPC. The CHPC will require bandwidth of at least 10Mb/s for national, and at least 155Mb/s for international, connectivity. The high cost of bandwidth is seen as a significant regulatory constraint, which currently inhibits the development of particularly networked high performance computing.

It is hoped that national high-bandwidth connectivity between CHPC member institutions across the country can be realised through the government's SANReN (South African National Research Network) initiative. The CHPC will also strive to provide South African researchers with the necessary international connectivity to participate in, benefit from, and contribute to global research activities. An agreement or partnership with Telkom, the national telecommunications company, leading to a donation (or at least an affordable supply) of the required national and international connectivity, will be negotiated.

6. Progress to date and the way ahead

Funding, at this stage largely from government, has been secured for the establishment of the central physical facility together with the appointment of scientific and technical staff by mid-2006. This major milestone has been preceded by a period of intensive planning extending over more than two years, in which stakeholders such as senior representatives of universities, the research community, relevant industries, and members of national government departments have worked to construct a common vision, strategy, and plan for operationalisation. In this regard, a key advisory role has been played by international colleagues with expertise in the establishment of HPCs. Linkages with similar facilities in developing countries such as Brazil and India are seen as essential to the success of the South African project, given this country's largely developing economy. And in this regard discussions have been held with CDAC in India with a view to establishing a relationship similar to that envisaged with LNCC in Brazil.

A key objective will be that of identifying projects that will be supported through the CHPC. These will be identified through a dual process of solicitation of proposals on the one hand and identification of project areas, typically of national importance, on the other, which are deemed to be appropriate for location in the CHPC. Project areas that are of interest include those in materials modelling and minerals processing and computational fluid dynamics with potential impact on mining and materials-related industries; bioinformatics and medical imaging technologies with impact on the health and pharmaceutical sector; geophysics with impact on the oil exploration industry; computational chemistry, drug discovery and design, HIV/AIDS research, molecular modelling to improve process mining; climate systems modelling, which can be used to study climate change and for disaster forecasting; radio-astronomy and astrophysics with particular reference to the SKA project; image and visualisation technologies which impact the film making and tourism industries; and defence applications such as radar detection and advanced weaponry development.

In this millennium we will see the use of computers become critical to problems as diverse as drug design to combat diseases malaria and HIV/AIDS through the development of models for predicting drought and preventing crop failures. High performance computing is now being positioned at the centre of innovative technologies. The impact of design through scientific computing on economies driven by innovation will be significant.

The creation of a national Centre for High Performance Computing will permit South African scientists and engineers to be active at the cutting edge of their respective research disciplines within a vibrant intellectual atmosphere. The benefits of the linkage between research and innovation that is enabled through the CHPC will be felt not only in university laboratories but throughout the wider South African economy. The building of a critical mass in state-of-the-art high-performance computing equipment as well as high-level scientific computing expertise in an intellectual common space will be central to achieving the goal of making the African Renaissance a reality.

1 http://www.chpc.org.za/
2 http://www.nepad.org/
3 http://www.meraka.org.za/

URL to article: http://www.ctwatch.org/quarterly/articles/2006/02/high-performance-computing-in-south-africa-computing-in-support-of-african-development/