Answers are grounded in this paper and cite their sources.
Investigates online scheduling of DAG tasks in multi-edge environments, suggesting improved efficiency.
With the rapid growth of data volume and increasingly stringent latency requirements in data-intensive real-time services within digital infrastructure, mobile edge computing (MEC) must process complex workloads efficiently under constrained computation and bandwidth resources. In practice, many workloads can be modeled as directed acyclic graphs (DAGs), where subtasks are subject to strict precedence constraints. Meanwhile, cooperative multi-edge environments feature heterogeneous computing capabilities, dynamic queue backlogs, and stochastic arrivals, making offline or static heuristics difficult to adapt. This paper investigates the online scheduling and offloading of dependency-constrained DAG tasks in a cooperative multi-edge setting, aiming to minimize the long-term average makespan, and proves that the resulting optimization problem is NP-hard. To this end, we propose an online cooperative scheduling framework, termed DRL-TCES. Specifically, DRL-TCES leverages a Top-k sampling based Deep Q-Network (DQN) to dynamically select a cooperative edge set, and then determines the workload partition ratios by enforcing completion-time balancing across selected nodes, jointly accounting for transmission, queuing, and computation overheads to enhance parallel execution efficiency. Extensive simulations show that DRL-TCES reduces makespan by up to 63.9% and achieves a 98.5% task success rate over strong baselines, demonstrating superior scalability and robustness.
No takes yet. Share an insight, caveat, or question.
Wu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: