Evaluating the Efficacy of LLM-Based Reasoning for Multiobjective HPC Job Scheduling

Prachi Jadhav, Hongwei Jin, Ewa Deelman, Prasanna Balaprakash

Abstract:

High-Performance Computing (HPC) job scheduling involves bal-
ancing conflicting objectives such as minimizing makespan, reduc-
ing wait times, optimizing resource use, and ensuring fairness. Tra-
ditional methods, including heuristic-based, e.g., First-Come-First-
Served(FJFS) and Shortest Job First (SJF), or intensive optimization
techniques, often lack adaptability to dynamic workloads and, more
importantly, cannot simultaneously optimize multiple objectives
in HPC systems. To address this, we propose a novel Large Lan-
guage Model (LLM)-based scheduler using a ReAct-style framework
(Reason + Act), enabling iterative, interpretable decision-making.
The system incorporates a scratchpad memory to track scheduling
history and refine decisions via natural language feedback, while a
constraint enforcement module ensures feasibility and safety. We
evaluate our approach using OpenAI’s O4-Mini and Anthropic’s
Claude 3.7 across seven real-world HPC workload scenarios, includ-
ing heterogeneous mixes, bursty patterns, and adversarial cases etc.
Comparisons against FCFS, SJF, and Google OR-Tools (on 10 to 100
jobs) reveal that LLM-based scheduling effectively balances multi-
ple objectives while offering transparent reasoning through natu-
ral language traces. The method excels in constraint satisfaction
and adapts to diverse workloads without domain-specific training.
However, a trade-off between reasoning quality and computational
overhead challenges real-time deployment. This work presents
the first comprehensive study of reasoning-capable LLMs for HPC
scheduling, demonstrating their potential to handle multiobjective
optimization while highlighting limitations in computational ef-
ficiency. The findings provide insights into leveraging advanced
language models for complex scheduling problems in dynamic HPC
environments.