EcoHPC: Energy-Conscious High Performance Computing

Treats power as a schedulable resource on heterogeneous HPC systems, developing recommendation-driven scheduling, multi-objective optimization, and adaptive runtimes to cut wasted energy without sacrificing performance.

As high-performance computing systems grow more heterogeneous, mixing CPUs, GPUs, and other accelerators, mismatched or idle hardware wastes power even as workloads deliver strong performance. EcoHPC treats power as a resource that can be scheduled and optimized alongside performance: recommendation techniques merge offline analysis with live profiling to inform scheduling decisions, job placement is formulated as multi-objective optimization across heterogeneous nodes, and adaptive runtimes anticipate an application’s changing needs to cut wasted power without sacrificing throughput. The project also supports training the next generation of HPC researchers and practitioners in energy-aware system design.

Team:

  • Students
    • Giacomo Brunnetta [PhD Student with Zhiling Lan]
    • Zhong Zheng [PhD Student co-advised with Zhiling Lan]
  • Faculty
    • Zhiling Lan (UIC) — PI
    • Michael E. Papka (UIC) — Co-PI

This work was supported by the National Science Foundation under Grant No. CCF-2515009 (SHF:Small: Beyond Performance: Energy-Conscious High Performance Computing). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.