Information Technology (IT) systems range from small embedded systems such as smartphones to large computer facilities such as data centers. Whether small or large, IT systems proportionately use lots of energy. They also waste a lot, as in shorter battery life or higher data-center operating costs, so improving their efficiency fulfills a critical national need. The objective of this project is to propose solutions to increase energy efficiency in IT systems by: 1) developing a well-trained student workforce, and 2) proposing new techniques and approaches capable of reducing energy consumption and losses. Students will be introduced to the research area of applying power optimization techniques to IT systems. The target populations are underrepresented students and those from institutions that do not offer PhDs. In addition to academic mentoring by Indiana University Purdue University Indianapolis (IU Indianapolis) faculty, this project includes mentors from industry.
Traditionally, IT curricula have been developed with disciplines that focus on parameters, such as performance and runtime optimization. Little or no attention has been devoted to achieving energy efficiency and consumption thresholds, especially from the software and algorithmic perspectives. This project will provide original experiences and singular opportunities for undergraduate students to work on innovative strategies to reduce energy consumption and losses in IT systems. In addition to the potential change in the career paths of the selected students, this project suggests how the electrical and computer engineering curricula can be impacted to incorporate energy efficiency in both hardware and software curricula.
This project will demonstrate novel methodologies in reducing energy consumption and losses within IT systems, which can be accomplished with a multidisciplinary approach. Topics will range from hardware to software solutions with efficiency as the main figure of merit. Traditionally, Multiple-Voltage Multiple-Frequency (MVMF) has been used to reduce energy consumption in computing systems. This project proposes MVMF2, which not only dynamically changes the computer clock frequency, but also dynamically changes power supply switching frequency for further loss reduction. While MVMF2 is considered a hardware optimization managed by software, the second research topic - a higher efficiency power supply based on a new power converter design - is purely a hardware solution. The third and fourth research topics are software solutions to reduce energy consumption in high memory demand applications and processing big matrix multiplication, respectively. The fifth research topic deals with energy consumption reduction techniques in wireless Internet of Things devices by modeling and optimizing power saving modes, such as reduced power operation, cyclical sleep, among others.
SEIRI Evaluator: Justin L Hess
Investigators: Euzeli dos Santos (PI); John Lee
Funding Organization: NSF Division of Computing and Communication Foundations
RFP: Research Experiences for Undergraduates
Grant Status: Current (02/01/2018 - 1/31/2021)
Award #: 1757636
Award Amount: $358,076
