Balkind – IEE Funding

Investing in Innovation

Three projects that take innovative approaches to solving critical energy-efficiency challenges have been awarded seed funding from UC Santa Barbara’s Institute for Energy Efficiency (IEE), an interdisciplinary research institute committed to improving energy efficiency. The selected projects align with at least one of the institute’s three key interdisciplinary thrusts: smart societal infrastructure, computing and communications, and the food-energy-water nexus. Each proposal will receive up to $50,000 in critical seed funding, which is intended to produce preliminary results that the scientists can use to apply for major external funding to expand their research. 

“Supporting projects in the early stages of development is an essential step to the creation and delivery of high-impact solutions to improve efficiency,” said IEE director, John Bowers, a distinguished professor of electrical and computer engineering (ECE) and materials. “Each project attacks a grand challenge with a leading-edge approach that has tremendous potential. They also foster new research collaborations and leverage IEE’s legacy of scientific discovery.”

Along with Balkind, the awards bring an additional five new faculty members from the departments of electrical & computer engineering, mechanical engineering, and physics to the IEE.

“The review committee was thrilled with these projects because they proposed innovative paths to discovery and enrolled new junior and senior faculty in the institute,” said Mark Abel, associate director and executive advisor of IEE. “We also saw how seed funding could really help these faculty move their new efforts forward, and how the research, if successful, could lead to additional funding opportunities.”

The selected projects are aimed at developing novel technologies, ranging from new computer hardware to a probabilistic computer, to an energy harvester that powers portable and remote smart devices. 

Microarchitectural Checkpointing 

Computer processors — digital circuits that perform operations on external data sources — have become more complex over the years, as their performance and efficiency have improved dramatically. One way that computer architects have increased their efficiency is by exploiting the predictability or “locality” of applications, designing them to save information, or learn as they run. Therefore, the longer an application runs, the more efficiently the processor can execute it. On the flip side, short-running applications are often significantly less efficient because processors cannot learn as much and are forced to start from scratch each time they are activated. Jonathan Balkind, an assistant professor of computer science, has received seed funding to improve the efficiency of short-term applications by focusing on building mechanisms, which he calls microarchitectural checkpoints, into the processor. 

“The checkpoints enable us to save what processors have learned for later in order to make these applications more efficient,” said Balkind, who joined UCSB in July 2020, after completing his PhD in computer science at Princeton. “The checkpoints would include what data is being frequently used or what paths of execution the program is likely to take. They would save that information when the program is shut down, so that when it is started up again, that checkpoint would load and, we hope, run the application far more efficiently by exploiting what was learned the last time it ran.”

The checkpoints could improve the efficiency of two major computing domains that rely on short-term applications: intermittent computing and Function-as-a-Service (FaaS). Intermittent computing systems perform as much computation as they can out of a limited energy supply, such as a solar-charged battery. These devices typically lose power while performing the work. Balkind proposes adding a new microarchitectural checkpointing mechanism into the hardware that can be customized to the particular application that will be run. 

“Since the processor may be turning on and off unpredictably, it’s difficult to build systems that can continue to execute useful applications and know if the data has been safely stored or whether it will work once the power returns,” explained Balkind. “There is a lot of research for these systems that we are interested in applying microarchitectural checkpointing to in order to see what benefits it can give us.”

FaaS is a type of serverless computing and a way to deploy applications in the cloud by loading functions and launching on demand. FaaS works well for stateless services, which rely only on information relayed with each request and do not rely on information from previous operations. However, the stateless and short-lived functions are problematic because they lose locality or reference, forcing the application to reload and process the information each time. Balkind’s previous research verified that FaaS applications were short running and could not predict or reference past information.  

“We propose adding mechanisms to the processor to save and restore the state of these locality-preserving structures, in order to start functions in a well-trained state, ready to operate with high efficiency,” said Balkind. “Realizing such capability will make FaaS a more predictable and efficient application that will be more attractive to developers.” 

Balkind says the project could lead to the design of  machines designed specifically for running short-lived applications like FaaS, as well as the rapid growth of energy-efficient IoT applications made possible by intermittent computing. 

Seed funding is supported by gifts from private donors, such as John MacFarlane, a member of the IEE Directors Council. Support from an anonymous donor allowed the IEE to award an additional grant this year. Those interested in providing seed funding should contact Mark Abel at markabel@iee.ucsb.edu.

Learn about the other awardees:

• Probabilistic Computer  – Kerem Camsari & Luke Theogarajan (Electrical & Computer Engineering)
• Miniature Energy Harvesters – Sumita Pennathur & Carl Meinhart (Mechanical Engineering) and David Weld (Physics)

COE News  – “Investing in Innovation" (full article)