Physics' new 20-PC CARMEN cluster a hit on the world stage, say all

21 October 2002
A TEAM of UCT physicists has received plenty of acclaim for their recently launched “supercomputer”—20 interlinked computers that run as one—and the system has already been ranked as number 55 on the list of fastest 100 such units in the world.

The cluster of computers, dubbed CARMEN by PhD student and group member Bruce Becker (reasons are manifold and labyrinthine, he observes), is the Department of Physics' contribution to the A Large Ion Collider Experiment (ALICE) of the European Organisation for Nuclear Research (CERN), the world's largest nuclear physics research centre, based in Geneva, Switzerland.

Over 78 institutes and about 1 000 scientists from 28 countries across the globe form part of the collaboration, which will try to recreate Quark Gluon Plasma (QGP), sometimes referred to as the “missing state of matter” (see Monday Paper, vol 21 no 10, April 29 to May 6, 2002). Becker and Horner set up the cluster at UCT.

CARMEN, like similar systems at major research centres around the world, will be used to analyse the vast amounts of data generated by ALICE. The UCT researchers who will work on the system are Professor Jean Cleymans, who heads the UCT Department of Physics and the UCT team; Dr Zeblon Vilakazi; Dr Roger Fearick; PhD students Becker, Mark Horner, Mark Marais and Spencer Wheaton; Masters student Nawahl Razak, and Honours students Heather Gray and Artur Szostak.

Praise came thick and fast at CARMEN's recent launch, attended by representatives from the Department of Science and Technology (DST) (which has contributed funds towards the UCT project), as well as ALICE collaborators from Switzerland, India, the United States and Germany.

Professor Volker Lindenstruth of the University of Heidelberg in Germany, for one, noted that his institute's cluster consists of 80 computers, but is nowhere near as fast and powerful as the UCT version.

Cleymans pointed out that such a cluster is much cheaper—and faster—than the conventional single supercomputer, each computer “box” (or node, as it is known to group members) in the system costing only R8 000. He also noted that his purchase bemused a host of computer salespeople who could not understand why he would need “just the box” sans monitors or keyboards or other appurtenances.

CARMEN happens to be so powerful simply because it consists of 20 brand new Pentium 4, 2.80 GHz machines, while other centres have “outdated” two- or three-year old units, Cleymans explained modestly. “So it's more by accident than anything else that it's so fast.”

Cleymans also observed that the cluster will not only be used for ALICE, but also for a number of other UCT research ventures, including earth observations, and biological and medical imaging. “The project was started for, but is not limited to, the High Energy Physics community or a narrow field of physics, and we want other departments in the Science Faculty to use it as well,” he said.

The youthful make-up of the UCT group—its high number of postgraduate students—was another feature of the project that attracted much comment and praise from the international collaborators at the launch.

“There's been a global decline in the number of physics students, but recently we have seen a reversal of this trend at UCT, with encouraging increases in enrolments, particularly at the senior undergraduate and postgraduate levels,” commented Professor Daya Reddy, Dean of the Faculty of Science, among others.

One of these postgraduate students, the ever-iconoclastic Becker, has assigned each node an Italian name, in honour of either visiting students or opera characters, just to make sure that the whole pursuit of knowledge does not become too alienating or impersonal. So, in the morning, he greets "Antonio" and "Luciana" and "Babbetta" and "Francesca" and "Chiara" as they gather their resources and make their mark on the world of physics.

For more information on the project, please look at Becker's websites at or .

Creative Commons License This work is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.

Please view the republishing articles page for more information.