Making Computing Sustainable, With Help from NSF Grant

With research projects - including one that recently received a $1.3 million grant - and an upcoming course, Prof. Robert Soulé is looking at new ways to make computing more sustainable. 

Working with Prof. Noa Zilberman from Oxford University, Soulé has received a grant jointly funded by the United Kingdom’s Engineering and Physical Sciences Research Council (EPSRC) and United States’ National Science Foundation (NSF) for work that aims to reduce the energy consumption of computing.  Specifically, it sets its sights on computer networks, which consume an estimated one-and-a-half times the energy of all data centers, according to some reports. In contrast to other large scale computer infrastructures, accounting for the carbon emissions of the network is extremely hard. 

The project is designed to collect information about the power consumption of network devices, specifically the computer hardware involved in connecting users to computer networks. This includes switches, which connect different computers together, and the network interface cards in computers or servers that connect users to the network. Traditional computer networks try to optimize the paths to reduce latency and achieve the fastest response possible. 

“But what we're hypothesizing is that you could actually instead choose paths that would result in the lowest amount of power consumed, or maybe the greenest path,” said Soulé, associate professor of computer science & electrical engineering. “We want to measure how much power they are consuming and the quality of the power that they're consuming. For example, did they come from a green energy source? So we’re collecting the data that would allow you to make these informed decisions, and designing the network algorithms that would change a routing behavior in order to reduce the overall carbon footprint.”   

One possible way to do that is to develop systems that send computer traffic to a path that consumes energy from a green energy source. Another is a system that chooses a path that minimizes overall power consumption. 

Another component to Soulé’s work in this area is a collaboration with Prof. Rajit Manohar, the John C. Malone Professor of Electrical Engineering and Computer Science. They’re developing network hardware that can go into idle mode when it’s inactive, very much like some cars with engines that automatically turn off at red lights. 

“There's a problem with current network hardware in that it’s not really able to go into idle mode because a part of it is always running to see if information is arriving,” he said. “So Rajit and I have been looking at whether we can design hardware devices - a new network switch - that would consume energy in proportion to the amount of traffic that it's seeing. And if it did see less traffic, it would go into idle mode” 

Soulé is also co-teaching a course next year on sustainable computing with Dr. Eve Schooler, an IEEE Fellow and Yale alum. The course, they said, takes a broader view of the subject. 

“We're trying to do more of a survey of different approaches to improving the carbon efficiency of computer networks in general,” Soulé said. “But even beyond that, we're also looking at a broader discussion at the policy level, where the intersection of sustainability and technology meet.”

Schooler said the course will cover a “large swath of topics.” For instance, it might explore issues like the role that computing can play in the Intergovernmental Panel on Climate Change or how large institutions perform carbon accounting. 

“We'll also focus on networking, and on topics around the streaming infrastructure, the content distribution networks,” she said. “Other topics will be about large algorithms, like large language models - the ChatGPTs of the world - and Bitcoin or some of the crypto currencies that are also large consumers of electricity.”