Collaborative Research: CSR: Medium: Offloading Heterogeneous Distributed Workloads with Diverse Computation Models

The ever-growing diversity of edge devices, from CPUs for basic tasks to graphics processing units (GPUs) for graphics and neural processing units (NPUs) for machine learning, presents a challenge for edge cloud computing. While advanced wireless communication seamlessly connects billions of edge devices to the edge cloud, traditional homogenous edge cloud platforms struggle to handle diverse workloads and computation models efficiently.

This research proposes a pioneering approach using Field-Programmable Gate Arrays (FPGAs) within an edge overlay framework. FPGAs can be dynamically reconfigured to act as various processing units, efficiently handling these diverse computational needs. The main goal of this research is to develop novel techniques for offloading heterogeneous tasks, ensuring high overall throughput, uninterrupted service, and fault tolerance.

To demonstrate the effectiveness of the proposed techniques, this research will focus on a drone network surveillance use case. The developed approach has the potential to significantly improve edge computing's energy efficiency, resiliency, and scalability.

This research will make a significant contribution by making powerful edge cloud computing more accessible. To achieve this, the researchers will develop new course modules at UMass and WPI focused on heterogeneous edge computing, institute a research workshop for sharing research ideas and showcasing work, and leverage targeted programs to recruit underrepresented students to research programs.

These initiatives will empower undergraduate and graduate students to leverage edge cloud FPGA resources for various hardware and software experiments. The annual research workshop, organized and executed by graduate students, will be open to the wider community, further expanding the project's reach and impact. All findings, innovations, and developed software from this research will be openly shared to ensure they are freely accessible and usable by the research community, industry partners, and the public, promoting collaboration, further development, and practical applications.

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.