It all began with parallel thinking. Parallel thinking led to parallel computing, which, as of the late 1990s in the Calvin computer science department, where this tale is set, had not yet come of age. Parallel computing is the simultaneous running of programs on multiple computers linked together. It is supercomputing. And Calvin had no supercomputer then.
Calvin did have a parallel computing class: “Students could write a program to run on a supercomputer, but it would run on a regular computer,” said computer science professor Joel Adams, who taught the class.
To purchase a supercomputer from a vendor of such wonders like Cray or IBM was, and remains, beyond the department’s means, but Adams was hearing rumors about NASA scientists building things they had named Beowulf clusters: supercomputers built of standard PCs and named for the epic hero who freed the Danish warriors from the monster Grendel. “They saw themselves as being like Beowulf and freeing researchers from the tyranny of vendors everywhere,” Adams explained the NASA moniker.
It was not an epic hero, however, but an enterprising student of computer science and employee of Calvin Information Technology (CIT), Mark Ryken ’99, who led the way into the parallel world. Ryken salvaged eight of the older-model computers that CIT was then jettisoning in favor of Pentiums and linked them, supercomputer-style. It was his senior project. The result, built with outmoded technology, wasn’t very fast, but it was an impressive feat, Adams said: “This was, as far as I know, the first time an undergraduate student had built one. All the others were built by research labs, grad students and PhDs.”
“I would hardly call what I built a supercomputer, but I suppose it was proof of concept,” Ryken recently demurred. “If one fairly average guy could put the pieces together on his own, a group of above-average students with some exceptional leadership should be able to do some pretty impressive work.”
Indeed, Ryken’s project served as the inspiration for the department’s first real supercomputer, which was funded in 2000 through a $163,000 grant from the National Science Foundation’s (NSF) Major Research Instrument program. And on Jan. 10, 2001, computer science professors, students and alums, and a staff member or two came together to build the first Calvin supercomputer. They brought in pizza and connected 18 computers via hypercube, a way of connecting computers. They called the result “Ohm” for “our hypercube multiprocessor.”
Ohm has served multiple uses. It is a teaching tool. “Now the class that inspired Mark Ryken in 1998 had an actual supercomputer to use,” Adams commented. The supercomputer is also a research tool for faculty from various disciplines. Ohm is proficient at research because every one of its parts can tackle a separate part of a complex problem.
“Ohm can allow us to do calculations in one evening that would take us a month on a regular computer.” — Paul Harper
Paul Harper, a Calvin professor of physics and astronomy who uses Ohm to model the transport mechanisms across lipid membranes, testified to the supercomputer’s superpowers: “Ohm can allow us to do calculations in one evening that would take us a month on a regular computer,” he said. Ohm’s third use is Ohm itself because Adams’ research is “topology”—the study of how to link the computers in a supercomputer. The star, the hypercube and the ring are three common computer topologies, and as a multi-topology computer, Ohm boasts all three.
What sets the Calvin supercomputer apart is its ability to run in any combination of star, hypercube and ring, a versatility Ohm owes to yet another intrepid undergrad, Kevin deGraaf ’03. It was deGraaf who devised the software that allows the computer to switch topologies without rebooting—another supercomputer first.
Yet, as talented as Ohm is, it is no longer state-of-the-art in the age of multiple processors. “As desktop computers get faster and faster, our supercomputer is staying the same speed,” Adams said. “It’s still useful for teaching, but less and less useful for research.” And in the age when desktop models sport even quadruple processors, students need to learn parallel thinking. “They need to think in parallel because all the hardware is going parallel,” Adams said.
Recently, the computer science department landed a second NSF grant for $205,000 to build a worthy successor to Ohm. The new supercomputer will be built of at least 32 computers, each with multiple cores, and it should be a pretty capable machine. “The new Ohm could run up to 100 times faster than the original,” Adams said. “We won’t know until we build it.”
While between Ohms, Adams built another supercomputer for the engineering school of Reykjavik University while he was teaching there on a Fulbright fellowship. That system, named Sleipnir for the Norse god Odin’s eight-legged dog, consists of six computers in a star topology. “They didn’t have one, so …,” he explained that endeavor.
Adams also partnered last year with student Tim Brom ’07 on Microwulf, a supercomputer measuring all of 11 inches by 12 inches by 17 inches. Microwulf, more than twice as fast as Ohm, is the most cost-efficient supercomputer anywhere that Adams knows of.
And, before long, he’ll be helping to build the new Ohm. “We’ll bring in pizza,” he said.
Giving to Calvin
Majors & Minors
People at Calvin