Cambridge, MASS. — The Massachusetts Institute of Technology’s Laboratory for Nuclear Science (LNS) announced a research collaboration with Compaq Computer Corp. of Houston to develop a cluster of extremely powerful computers capable of doing calculations that will help researchers understand the structure of subatomic particles. Calculating the structure of the proton from first principles requires a computer that can achieve one teraflops – the ability to compute 1 trillion floating point operations in one second.
Using a conventional supercomputer for this calculation could cost up to $500 million. A large special purpose computer would cost approximately $10 million per sustained teraflops, but would lack the programming flexibility needed by physics researchers.
Under the research agreement, LNS has purchased a 64 gigaflops prototype cluster of 12 Compaq AlphaServer ES40 systems and will collaborate with Compaq to optimize its performance for fundamental physics calculations.
“The goal of the MIT-Compaq collaboration is to create a network of AlphaServer computers running the Opn Source Linux operating system that will be far more flexible than special purpose machines at a much lower cost than conventional supercomputers,” said Bill Blake, Compaq’s vice president for high performance technical computing. “With Alpha’s exceptional floating point performance and high memory bandwidth, it is the processor of choice for such high-performance scientific clusters.”
The cluster will be used for research in quantum chromodynamics (QCD) – which describes how protons and other strongly interacting particles are built out of quarks and gluons, the basic building blocks of matter – and related computationally intensive science problems.
Robert L. Jaffe, professor of physics and director of MIT’s Center for Theoretical Physics (CTP), said, “The computational approach to QCD represents one of the most promising means of unraveling this puzzling and fascinating theory. This field is certainly only in its infancy, with many dramatic achievements lying ahead. We are very pleased to have this world-class initiative embedded within the CTP, where we can share insights with some of the world’s finest computational physicists.”
The discovery of the quarks as the fundamental constituents of the proton 30 years ago by Nobel laureates Jerome Friedman and Henry Kendall of MIT and Richard Taylor of the Stanford Linear Accelerator Center laid the foundation for quantum chromodynamics.
QCD describes how protons and other strongly interacting particles that comprise most of the known mass of the universe are built out of quarks and gluons. The only way to solve QCD is by numerical computation, and with the advent of terascale computers, it is now possible to calculate the quark and gluon structure of matter from first principles.
The MIT-Compaq initiative, led by nuclear theorist John W. Negele, William A. Coolidge Professor of Physics at MIT, has as its next step a proposal by MIT and Jefferson National Accelerator Facility (JLab) in Newport News, VA, to add two additional large Alpha clusters to bring the total resources of the two institutions to half a teraflops.
Based on the success of this half-teraflops project, the next phase would be a multi-teraflops cluster based on Compaq’s EV-7 technology, which places all the essential functions of the present Alpha system on a single chip.
A team of 22 physicists from 13 institutions in the Lattice Hadron Physics Collaboration will use the MIT and JLab clusters to calculate in detail the quark structure of the proton.
“Research in theoretical nuclear physics always has been one of the strengths of the Laboratory for Nuclear Science. This new initiative will allow LNS to remain at the forefront of this field,” said June Matthews, director of the LNS.