Conserving energy improves performance and helps the environment Hero Image

Increasing the efficiency of everything from cell phones to supercomputers

Daniel Mosse, Professor

U.S. data centers consumed approximately 70 billion kilowatts of electricity in 2014. That’s a little more than the combined amount used that same year in Ireland and Hong Kong. Computers are ubiquitous and our dependency on them will continue to grow. Is it possible to balance our desire for fast computing while simultaneously reducing the amount of energy used? Can changes be made — to battery operated devices, supercomputers, and our heating systems — enabling them to use energy more efficiently without compromising user expectations?

The research by Daniel Mosse focuses on reducing the amount of energy used in computing systems.

For as long as he can remember, Daniel Mosse has been concerned about being efficient. Therefore, it’s no surprise that his research deals with how to direct energy to increase efficiency. When computers are designed to operate more efficiently, it has a positive impact on the environment, plus efficient power management can extend the life of devices since using less power means fewer equipment failures.

One of his research projects looks at managing energy consumption in devices that are battery operated, such as phones and laptop computers.

Most computers have multiple central processing units (CPUs). Mosse and his colleagues are researching the benefits of having a small CPU and a large CPU inside the computer and enabling the device to automatically switch between the CPUs based on the amount of energy needed. For example, reading emails could be done on the small CPU, if you’re playing a video game, the computer would automatically switch to the large CPU.

Another one of his research projects looks at saving energy on a much larger scale.

U.S. supercomputers are used as a computational tool to solve some of the world’s most complex problems. To do so, they use an enormous amount of energy. For example, the Titan supercomputer at the Department of Energy’s Oak Ridge National Laboratory uses approximately the same amount of electricity as the amount used by 9,000 homes.

The U.S. Department of Energy has a mandate that their next supercomputer can only use up to 20 megawatts, yet they want it to be 1000 times faster. Mosse is researching ways to reduce the time the CPUs in supercomputers sit idle, which is simply wasting energy. Even if Mosse can find a way to save 10 percent of the current energy usage, it would add up to a substantial savings.

His interest in energy efficiency and resource allocation led to the creation of HiberSense, a startup company he founded with two alumni of the University of Pittsburgh Swanson School of Engineering, Jacob Kring and Brendan Quay.

HiberSense is developing a smart thermostat system that monitors temperature, light, motion, humidity, and sound to adjust heating and cooling equipment accordingly. The company was one of 35 companies from across the nation selected to participate in the first-ever University Startups Demo Day in Washington D.C. in 2016. In addition, they were one of the winners in the 2017 UpPrize competition, a social innovation challenge that was created through a partnership of BNY Mellon, the BNY Mellon Foundation of Southwestern Pennsylvania, and The Forbes Funds

The system uses machine learning and control algorithms to anticipate the desired temperature and wireless, motorized vent covers that automatically direct airflow to only the rooms that need it, when they need it.

“HiberSense is a good example of solving a problem in context,” adds Mosse. “As a consumer of heating, I saw that there was a problem, but it’s not a computer science research problem, it’s an application of existing computer science and engineering solutions.”


Research Interests:

Power, energy and temperature aware systems, real-time systems, scheduling, distributed systems, wireless networking


“If you can make something efficient and save the environment at the same time, I’m all for it,”

Mosse

Daniel Mosse is a professor at the University of Pittsburgh’s School of Computing and Information.