GrapheonRL: A Graph Neural Network and Reinforcement Learning Framework for Constraint and Data-Aware Workflow Mapping and Scheduling in Heterogeneous HPC Systems

2025-05-30

AI Summary

This work addresses the challenge of mapping and scheduling large-scale scientific workflows in heterogeneous HPC environments under dynamic constraints and data-awareness requirements. We propose an end-to-end, constraint-aware scheduling framework that tightly integrates Graph Neural Networks (GNNs) with Proximal Policy Optimization (PPO). Workflows are modeled as directed acyclic graphs (DAGs), where GNNs jointly encode task dependencies, resource heterogeneity, and hard constraints (e.g., deadlines, data locality), while PPO enables adaptive, search-free online scheduling via reinforcement learning. Evaluated across multiple real-world datasets, our approach achieves 76% speedup over ILP-based solvers while maintaining near-optimal solution quality, and incurs only a 3.85× runtime overhead compared to the OLB heuristic. The framework seamlessly integrates with SLURM and Kubernetes, strictly satisfies all hard constraints, and demonstrates strong generalization across diverse workflow and infrastructure configurations—effectively balancing optimality and real-time responsiveness.

Technology Category

• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Machine Learning: Hardware-aware ML

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Effective resource utilization and decreased makespan in heterogeneous High Performance Computing (HPC) environments are key benefits of workload mapping and scheduling. Tools such as Snakemake, a workflow management solution, employ Integer Linear Programming (ILP) and heuristic techniques to deploy workflows in various HPC environments like SLURM (Simple Linux Utility for Resource Management) or Kubernetes. Its scheduler factors in workflow task dependencies, resource requirements, and individual task data sizes before system deployment. ILP offers optimal solutions respecting constraints, but only for smaller workflows. Meanwhile, meta-heuristics and heuristics offer faster, though suboptimal, makespan. As problem sizes, system constraints, and complexities evolve, maintaining these schedulers becomes challenging. In this study, we propose a novel solution that integrates Graph Neural Network (GNN) and Reinforcement Learning (RL) to flexibly handle workflows, dynamic constraints, and heterogeneous resources while providing quick responses. GNN manages dependencies and resource requirements, and RL optimizes scheduling decision-making via a learned policy, overcoming the need for a comprehensive global search. Experimental results with different datasets demonstrate that this method effectively adapts to different workflows, adheres to HPC constraints, and offers optimal solutions akin to ILP but with drastically reduced execution times (76 percent faster), comparable to heuristic methods (only 3.85 times slower than OLB). Our contribution is to provide a robust yet scalable mapping and scheduling solution that can handle changing constraints, as well as workload sizes and complexities in a heterogeneous HPC Compute Continuum system landscape.

Problem

• Optimizing workflow mapping in heterogeneous HPC systems
• Handling dynamic constraints and resource requirements efficiently
• Reducing makespan while maintaining optimal scheduling solutions

Innovation

• Integrates GNN and RL for workflow scheduling
• Uses GNN for dependencies and resource management
• RL optimizes decisions with learned policy