High-Performance Computing (HPC) is essential for advancing fields like climate modeling, cancer genomics, and AI training, yet modern large-scale HPC systems are vastly underutilized, with only 3% of computing power effectively used for typical workloads. This inefficiency not only wastes money and energy but also hampers innovation. A key issue lies in the communication-heavy demands of current applications, which create bottlenecks in HPC networks.

This project seeks to address this challenge by focusing on network-oriented resource management to boost HPC efficiency. Unlike traditional methods that focus solely on compute resources and static behaviors, we will develop innovative strategies for monitoring and modeling large-scale networks. These insights will inform scalable solutions for managing both compute and network resources in HPC environments. Additionally, the project will explore dynamic job scheduling, unlocking new opportunities to optimize system performance.