Optimal Multi-Constrained Workflow Scheduling for Cyber-Physical Systems in the Edge-Cloud Continuum

TL;DR

This work tackles latency-critical workflows in edge-hub-cloud CPS with heterogeneous, resource-constrained devices. It introduces an offline continuous-time MILP formulated on an Extended Task Graph (ETG) that captures device capabilities, memory/storage budgets, energy consumption, and communication costs, aiming to minimize the completion time $T$ within a deadline $L_{ ext{thr}}$. The paper presents a comprehensive MILP model including candidate-node and arc-selection, precedence, non-overlap, and device-resource constraints, and compares it to an enhanced HEFT baseline extended to the same ETG, using a real UAV inspection workflow and synthetic scalability tests. Results show MILP consistently surpasses HEFT in latency (average improvements of $13.54\%$ on real-world and $33.03\%$ on synthetic workloads) and also reduces energy consumption, with solver runtimes in a practical offline range, demonstrating both accuracy and scalability for edge-hub-cloud CPS scheduling.

Abstract

The emerging edge-hub-cloud paradigm has enabled the development of innovative latency-critical cyber-physical applications in the edge-cloud continuum. However, this paradigm poses multiple challenges due to the heterogeneity of the devices at the edge of the network, their limited computational, communication, and energy capacities, as well as their different sensing and actuating capabilities. To address these issues, we propose an optimal scheduling approach to minimize the overall latency of a workflow application in an edge-hub-cloud cyber-physical system. We consider multiple edge devices cooperating with a hub device and a cloud server. All devices feature heterogeneous multicore processors and various sensing, actuating, or other specialized capabilities. We present a comprehensive formulation based on continuous-time mixed integer linear programming, encapsulating multiple constraints often overlooked by existing approaches. We conduct a comparative experimental evaluation between our method and a well-established and effective scheduling heuristic, which we enhanced to consider the constraints of the specific problem. The results reveal that our technique outperforms the heuristic, achieving an average latency improvement of 13.54% in a relevant real-world use case, under varied system configurations. In addition, the results demonstrate the scalability of our method under synthetic workflows of varying sizes, attaining a 33.03% average latency decrease compared to the heuristic.

Figures (7)

• Figure 1: Transformation example: (a) the initial TG and (b) the final ETG.
• Figure 2: System model.
• Figure 3: Real-World Workflow Tasks
• Figure 4: Comparative evaluation between proposed MILP approach and enhanced version of HEFT for the real-world workflow under all system configurations.
• Figure 5: Latency decrease achieved by proposed MILP approach over extended version of HEFT under increasing TG size. Box plot in red shows overall distribution of latency decrease across all TG sizes.