Optimization of Workload Mapping and Scheduling in Scaling Heterogeneous HPC Systems: A Systematic Literature Review

Abstract

The increasing heterogeneity of High-Performance Computing (HPC) systems, incorporating Central Processing Units (CPUs), Graphics Processing Units (GPUs), Field-Programmable Gate Arrays (FPGAs), and accelerators, makes workload mapping and scheduling a critical optimization challenge. We present a systematic literature review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, covering 202 studies published from 2017 to 2026, and classifying optimization techniques for workload mapping and scheduling in heterogeneous HPC environments. We identify four solver classes: mathematical programming, heuristics, metaheuristics, and emerging methods (ML/AI, hybrid, and quantum). Makespan minimization is the dominant objective and multi-objective formulations are near-universal; ML/AI approaches have grown from negligible to dominating recent publications, while quantum methods are concentrated almost entirely in 2025-2026. Critical gaps constrain practical relevance: most studies address scheduling in isolation without joint mapping treatment; evaluations are predominantly small-scale and simulation-based; real-system evidence is sparse; and openly reproducible implementations are rare. These findings signal that the field's pressing challenge is not algorithmic diversity but deployment-grade evidence: solutions jointly formulated, validated at production scale, and openly reproducible.

The report follows IEEE double column format. It is accepted and presented at IARIA - SCALABILITY Conference 2026 on September 27, 2026 Contribution Id: "20007". Reprints and permission: https://www.iaria.org/conferences2026/SCALABILITY26.html