Blue Gene Sparks Research into Quarks
By Hannah Hsieh
It is MIT’s most powerful computer, with a theoretical peak speed of 5.7 Teraflops (5.7 x 1012 floating point operations per second) and sustained speed of 4.7 Teraflops. It is currently ranked 76th on the Top 500 list of computers in the world, of which most are other Blue Gene/L computers at other institutions.
A member of a new line of IBM supercomputers, MIT’s new Blue Gene computer consists of a single rack with 1,000 processor chips. They were developed through a partnership with the Lawrence Livermore National Laboratory in California, which currently owns 64 of its own. These computers represent a new generation of computing, in which a single supercomputer rack can replace a room full of conventional computers and use an order of magnitude less power.
The MIT computer is being put to work solving extremely demanding physics simulations, such as work by Professor John W. Negele on lattice quantum chromodynamics. QCD is the theory governing strong interactions between quarks and gluons, the fundamental units that comprise protons and neutrons. MIT’s Blue Gene computer is subsidized by the Department of Energy (DOE).
Coming to understand the properties of quarks revolutionized our understanding of nature in the twentieth century. MIT Professor Frank Wilczek shared the 2004 Nobel Prize in Physics for discovering in the early 1970s the property of asymptotic freedom, which predicts that the interactions between quarks are weak at short distances but peculiarly become extremely strong at large distances. Because of this property, termed confinement, it would take an infinite amount of energy to separate two quarks.
The only known way to calculate the properties of these strongly interacting particles is through lattice QCD, which formulates the theory of quarks and gluons onto a lattice that carves space-time up into discrete chunks. To predict how the quarks will behave over time, the supercomputer calculates the probabilities of all possible paths a quark could take through the lattice and uses them to determine the next state of the system in time.
The Blue Gene’s architecture is especially efficient for lattice QCD computing because it splits up the lattice onto separate processor chips. Instead of having one chip calculate the entire lattice, which could take years to compute, the work is distributed among 1,000 chips.
Negele said that the basic goal of the simulations is to ask the question, “How do those simple interactions give rise to more complex structures?” Most basically, particles like protons and neutrons are composed of different combinations of quarks, and the nucleus of an atom may contain tens of protons and neutrons, all in a structure often stable enough to last thousands of years without breaking apart.
For example, Negele’s team recently published a paper in which they computed a governing parameter for how a neutron decays to a proton (called neutron beta decay) to 7 percent precision using the new Blue Gene computer. Higher precision predictions using simulations of a theory allow for a more careful comparison of experiment and the theory.
The QCD team includes research scientist Andrew Pochinsky, who is working on optimizing the computing software, graduate student Dmitry Sigaev, and undergraduate Patrick S. Varilly who is working on his senior thesis exploring diquarks in protons.
While not being used for QCD, MIT’s Blue Gene is also part of a government research project. Negele said one reason the DOE chose MIT as a recipient of the Blue Gene computer is because the department hopes the quality of MIT’s engineers will prove useful in further optimizing the supercomputers.
MIT programmers work on optimizing performance and getting the computer up to maximum speed and also figure out what kind of software tricks can be used to develop tools for others. Blue Gene can also potentially be used for other areas of research including plasma physics, astrophysics, condensed matter physics, combustion physics, and ocean atmosphere modeling.
The computer is stored on the far west side of campus in Building W91, which is especially designed to support an uninterrupted power supply. Negele said the introduction of the Blue Gene computer was also beneficial for energy considerations at MIT because it “stimulated MIT to rethink what they’re doing about power supply.”
Since Negele hopes to bring another Blue Gene computer to MIT, “the institution has to supply the power and infrastructure” to accommodate another such powerful machine.
The supercomputer was dedicated on Nov. 10, according to a press release provided by Negele.