CCF Core: Small: Hardware/Software Co-Design for Sustainability at the Edge

Duke University, Durham NC

Investigators

Abstract

Due to the wide proliferation of cloud and now edge servers, the development of sustainable acceleration methods for edge and cloud computing and artificial intelligence (AI) acceleration is beneficial to society broadly and the computer science community, in particular. Reducing carbon and methane emissions is important to reaching environmental goals for reducing warming for which targets of 2030 and 2050 have been set. This project tackles the broader implications of the growing popularity of edge and cloud computing for executing AI to establish a sustainable computing system design methodology beyond pure energy efficiency. Unlike solely energy-efficient designs that focus on reducing power consumption and carbon emissions as a by-product, this project will build a hardware/software co-design framework to generate the optimal hardware and algorithm designs that meet specific constraints of functionality, performance, sustainability, and other system requirements. The framework will guide future sustainable hardware design. Towards this goal, the proposed framework engages holistic co-design efforts across four levels - modeling, algorithm, scheduling, and hardware. By introducing new synergies between computing paradigms and emerging AI applications, the outcomes of this research will benefit the entire AI industry, from hardware development to algorithm design and the applications for the end-users. The proposed co-design framework will be the first research in the community to build holistically sustainable AI systems which are recently recognized as emerging in importance. The success of this project will pave the road for future sustainable computing system design. The educational efforts aim at cultivating students' interests in the study of sustainable computing, contemporary computer architecture, and artificial intelligence. The existing curricula of computer organization, algorithms, and computing systems will be enhanced by the interdisciplinary research topics on sustainable computing, edge computing, and machine learning, as well as the hands-on experiences in building the simulation prototypes. Special attention will be given to recruiting underrepresented groups and enriching studentsí study experiences through new education forums. This project is funded by funds allocated to Design for Sustainable Computing (NSF 22-060) This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.