Supercomputers of the World: this one in Ghana

By Anna Bon

I never expected to find a supercomputer in Ghana, until I accidentally ran into one. My interest in globalization and information technology in relation to development made me visit the Kofi Annan Centre for Excellence in ICT, in Accra, a training centre for ICT personnel. My contact person was a gentleman named Mohammed who showed me around in the premises. Here, at the Kofi Annan Centre, is where the Ghana Internet Exchange is hosted. The exchange consists of a network switch where the glass fibre networks of local telecom companies are interconnected, enabling them to exchange their internet data traffic flows.

I was already surprised by this high-tech centre, when I suddenly became aware of a sign pointing to the HPC Department. As a former employee of the Dutch Supercomputer Centre, SARA, I am familiar with the concept of High Performance Computing (HPC) also called supercomputing. One does not come across supercomputers every day.

Supercomputers are not for common users. Supercomputing is only used for highly calculation intensive tasks, such as weather forecasting, molecular modeling, and nuclear physics. Besides, supercomputers are tremendously expensive and they require special operational skills for maintenance. Supercomputer owners are mainly universities, military centers and multinationals such as Shell. Why on earth, I asked myself, this centre here, was hosting a supercomputer.

The server room and rack were opened for me and I had the privilege of seeing the system. It was a Param Padma, designed and built in India by a centre called CDAC. The government of India donated this system to the Kofi Annan Centre.

Until that moment I had never thought other companies but the American giants IBM, Cray, SGI, Dell, or the Japanese NEC, or even the French Bull were into this HPC business. I could hardly believe my eyes. Supercomputing in Sub-Saharan Africa, using a system made in India.

Supercomputers have in common with pumpkins that they compete in size. A supercomputer competition is held every year for “benchmarking” the fastest system in the World. The 2007 TOP500 list shows IBM’s Blue Gene as the fastest supercomputer with a peak performance of 280 TeraFLOPs, which means it can make 280 trillion Floating Point OPerations per second. (A floating point is a number with a dot, such as 2.1; an operation can be e.g. 2.1 x 344.78). Because of their price and purpose most supercomputers of the World are located in the US and Europe.

As we all know, the US government, and especially the President Bush administration, is concerned about world security matters. They are worried some countries might use supercomputers for evil causes. Strict supercomputer export regulations have been set up to countries of “proliferation concern” which are countries with a nuclear program. Many countries, including India are not allowed to buy any supercomputers. India took this import restriction as a challenge and entered the era of supercomputer development. The Param Padma was launched in 2005, a machine originally consisting of 248 nodes, having a peak performance of 4 TeraFLOPs, i.e. four trillions floating point operations per second.

Now China is even more ambitious. China started a huge project to develop the world’s largest supercomputer, called Lenovo in 2005. This system will have a peak performance of about 1000 TeraFLOPs, which is more than 30 times as fast as the current world champion, the IBM Blue Gene. China’s economic growth is enormous. If they cannot buy a supercomputer, they will simply build their own system.
In our globalizing World technological bases are shifting from west to east, and probably some day from north to south, creating this Flat World, which was described by Thomas Friedman.

And I believe we should not underestimate the developing countries’ emerging innovative talent.