Supercomputers aren’t cheap, and it’s rarely easy to explain to people why $300 million should be spent on a facility that houses just a bunch of computers, hard drives, and massive air conditioning equipment. They also suck a lot of energy, which seems to run directly counter to calls for energy efficiency and conservation. The National Security Agency, for example, has created a mammoth 1.2 million square foot facility in the deserts of Utah that cost at least $1.5 billion and consumes 65 MW of power and 1.7 million gallons of water per day. That’s a lot of expense, and that’s just one computing facility. Are these kinds of projects worth it?
When I had a Visiting Research Scientist position at the University of Illinois at Urbana-Champaign a few years ago, I was intrigued and inspired by a project that was taking place there to build what, at the time, was designed to become “the world’s most powerful computer”. This NSF-funded project ran into some snafus, the most serious of which was IBM’s decision to pull out as the hardware developer for the project in 2011, which left the center with no actual computing equipment. After some delay, Cray Inc. stepped up as the hardware developer, and the center, called Blue Waters, began opening in stages in 2012.
At a cost of $300 million (of which the building itself was $72 million), Blue Waters ranks today as one of the world’s fastest computers. It performs 13.3 petaFLOPS (that’s a thousand trillion arithmetic operations involving decimal operations per second), has 1.5 petabytes worth of memory, and will eventually be able to store 500 petabytes worth of data. The machines are housed in 30 large cabinets separated by aisles of cooling ducts. These cabinets give off so much heat that the temperature differential between where the heat is forced out of the cabinets and the cooling air is sucked in is around 30 degrees. Yet, the facility has an energy efficiency in the range of 80% to 90%, which far exceeds the efficiency of the typical data center, which is only about 40%.
So, what is it used for? It’s not just a really expensive monument to silicon icons. It turns out that a lot of very interesting and promising science is being done at this facility. In one example, described in this video, researchers at Illinois are using Blue Waters to study the HIV virus. Specifically, they are using the facility’s massive computing power to simulate the interactions of 64 million atoms that form the HIV capsid, the protein shell that protects the virus’ genetic material. The hope is that, by modeling and visualizing how the atoms work together to protect the HIV genetic material, drugs can be designed to interfere with that cooperation, and the virus can then be neutralized. It is potentially high-impact work that involves mathematical operations in quantities that can only be accomplished by a machine with Blue Waters’ specs.
As is the case with space exploration, the field of supercomputing is unbelievably expensive and sometimes hard to justify in light of the many pressing priorities we face. But supercomputers are the only devices capable of performing the research for which they are being used, and some of the problems they are being asked to solve are among the most important of our time.