COVER STORY

CHIPPING AWAY AT BIG DATA’S BIG ELECTRIC BILL Mining Big Data guzzles energy. But Binghamton researchers are fine-tuning computer programs to run more efficiently. Information processing works in much the same way as an automobile assembly line. If one worker on the assembly line works faster than the rest, it doesn’t boost productivity; it gums it up. “Parts will back up and clog the system,” says Yu David Liu, an assistant professor of computer science. The solution is to tell the wayward worker to get in sync with the others. Computers work pretty much the same way. Let’s say Big Data processing software has 10 components, one of which is running too fast. Binghamton’s fix — a compiler optimization technique called Green Streams — is designed to figure out which one it is and make the necessary adjustment. Yu David Liu

Binghamton University • BINGHAMTON RESEARCH • Winter 2014

“We don’t want to slow down the overall system,” says Liu, an expert in energy-aware computing whose work has been funded by the National Science Foundation and Google. “We do want to figure out the best performance you can get, while saving as much energy as possible. We achieve this by slowing down individual components that run unnecessarily fast.” Liu and one of his doctoral students, Thomas Bartenstein, published their work on Green Streams this year in the 35th International Conference on Software Engineering, a premier conference in computer science. The project is good for the bottom line and for the environment. Lower energy use means lower power bills. And, says Liu, “You would produce fewer greenhouse gases to produce the same volume of work.”

18

When researchers get it right, their efforts produce a measurable impact on the hospital’s finances, but Khasawneh says there’s more to it than money. “It may help reduce hospital costs and improve profits,” he says. “But from my perspective, the most important thing is that it improves patient outcomes.” Hit the road

In order to build new roads, plan bus routes or organize the emergency evacuation of a large city, officials need to know where people go and where they congregate. That is a complicated matter. In New York City, for example, someone may be on the subway at rush hour, on Wall Street for the workday, at a Chinese restaurant for an hour after work and then at Yankee Stadium for a night game. After that comes a brief stop at a local watering hole and then a taxi ride home. When huge numbers of people are involved, Big Data can help sort out the details. Researchers in systems science and industrial engineering at Binghamton used anonymous social network data to analyze movement patterns in the city. The team — Associate Professor Sarah Lam, Assistant Professor Sang Won Yoon and graduate students